Anthropic Public Benefit Corporation and Long-Term Benefit Trust Explained

Anthropic is a Delaware public benefit corporation whose charter purpose is broad, whose directors are legally required to balance profit against that purpose, and whose public stockholders have a narrow, ownership-gated path to enforce that balance. The statute creates latitude for the board, not a private right of action for “humanity.”

Anthropic’s status and stated public benefit

Anthropic, PBC was incorporated in Delaware on January 26, 2021, and remains a public benefit corporation under DGCL Subchapter XV (Sections 361–368). Article III of the amended and restated certificate of incorporation filed as Exhibit 3.3 to its October 26, 2026 Form S-1 states: “The specific public benefit that the Corporation will promote is to responsibly develop and maintain advanced AI for the long-term benefit of humanity.” The same formulation appears in the S-1 prospectus summary, which calls this the “Company Purpose,” and in a May 7, 2026 amended certificate described by Reuters. [1] [2] [3]

That language is broader than the founding charter. Certificates obtained from Delaware and reviewed by Simon Willison show the 2021 original committed the company “to responsibly develop and maintain advanced AI for the cultural, social and technological improvement of humanity.” Later filings replaced that with the “long-term benefit of humanity” formulation. [4]

Under DGCL § 362, a PBC is still a for-profit corporation. It must identify one or more specific public benefits in its certificate and state in its heading that it is a public benefit corporation. “Public benefit” means a positive effect, or a reduction of negative effects, on persons, entities, communities, or interests other than stockholders in their capacity as stockholders, including effects of a scientific or technological nature. [5] Anthropic’s own description of the form matches the statute: PBC status requires it to balance stockholders’ financial interests with benefit to the broader public. [6] [7]

The charter purpose is not self-executing. In its 2026 Public Benefit Report, Anthropic says the purpose is “broad by design,” and that the board, with input from the co-founders and the Long-Term Benefit Trust, adopted five operational “charter pillars”: solving the technical problems of AI safety; making safety a universal, legally backstopped practice; strengthening democracy and global stability; expanding prosperity and human flourishing; and sharing AI’s benefits. Those pillars are a board framework for measuring the charter, not additional charter text. [8]

What this means for public stockholders. Buying Class A stock is a purchase of a for-profit issuer whose legally specified objective is not stockholder wealth. The purpose is so general—“long-term benefit of humanity”—that almost any safety delay, deployment restriction, or philanthropic allocation can be described as advancing it. The constraint is process (did the board balance?), not a judicially administrable outcome (did humanity benefit?).

The directors’ balancing duty under DGCL § 365

Section 365(a) is the operative duty. The board “shall manage or direct the business and affairs of the public benefit corporation in a manner that balances” three things: the pecuniary interests of the stockholders, the best interests of those materially affected by the corporation’s conduct, and the specific public benefit identified in the certificate. [5] Anthropic’s S-1 restates that duty and says PBC status “gives our board of directors the necessary latitude” to strike that balance, including weighing long- and short-term externalities of decisions such as whether to deploy a system. [2] [9]

The statute then strips most of the teeth out of that mandate:

• No duty to the beneficiaries. Section 365(b) provides that a director does not, by virtue of the public-benefit provisions or § 362(a), owe any duty to any person on account of that person’s interest in the chartered benefit or on account of being materially affected by the company’s conduct. The public, employees, customers, and “humanity” are not duty-holders. [5]
• Safe harbor. For a decision that implicates the balancing requirement, a director “will be deemed to satisfy” fiduciary duties to stockholders and the corporation if the decision is informed, disinterested, and “not such that no person of ordinary, sound judgment would approve.” That third prong is a waste-like standard. [5]
• Exculpation preserved. Under § 365(c), stock ownership alone is not a conflict on a balancing decision. Absent a conflict, failure to satisfy the balancing requirement is not, for purposes of the exculpation statute (§ 102(b)(7)) or indemnification (§ 145), an act or omission not in good faith or a breach of the duty of loyalty, unless the certificate says otherwise. [5]

This is the opposite of ordinary Delaware stockholder primacy. In a conventional corporation, directors may consider other constituencies only insofar as doing so advances stockholder pecuniary interests. In a PBC, those other interests are independent inputs the board must weigh, and a decision that sacrifices near-term value in service of them is not a breach merely because it does so. [10] [11]

Delaware’s first written decision on the point, Drakes Landing Associates, L.P. v. Tilden Park Capital Management, L.P., C.A. No. 2025-0898-NAC (Del. Ch. July 29, 2026) (Cook, V.C.), confirms that reading. In a rescue financing that would have diluted existing holders sharply, Vice Chancellor Cook held that Revlon’s exclusive stockholder-value-maximization mandate, read as a standard of conduct, cannot apply to a PBC: a singular duty to get the best price reasonably available cannot be squared with § 365(a). The court dismissed the complaint under the § 365(b) safe harbor. Plaintiffs had not pled facts supporting a reasonable inference that the special committee was uninformed, interested, or had approved a decision no person of ordinary, sound judgment would approve. The court left open whether a modified “PBC enhanced scrutiny” could still apply as a standard of review in a sale of control, but did not need to decide it. It also treated § 365(b) as negating the underlying breach, which knocked out an aiding-and-abetting claim against the financing parties. The opinion is a Court of Chancery dismissal, not a Delaware Supreme Court ruling, and it turned heavily on what the complaint failed to plead. [10] [12] [13]

The decision tracks commentary that has been on the books since the statute was young. Former Chief Justice Leo E. Strine, Jr. wrote that a consequential feature of the Delaware PBC statute is that it takes Revlon’s exclusive value-maximization objective off the table, and that a board’s good-faith balancing is protected by the business-judgment-style safe harbor in § 365(b). Making It Easier for Directors to “Do the Right Thing”?, 4 Harv. Bus. L. Rev. 235, 243 (2014), cited in the Drakes Landing commentary. [14] [15] Practitioner notes after the decision emphasize the practical lesson: document the three-way balance contemporaneously, because the safe harbor is a motion-to-dismiss defense, not a trial defense. [12]

Delaware corporate lawyer Lawrence Hamermesh’s assessment of what “balance” actually requires, given to CNBC in the context of Anthropic’s IPO, is blunt: if asked what the obligation to balance purpose and profit means, “I would say, ‘Good luck.’” [16]

What this means for public stockholders. A public holder cannot sue because the board chose safety over a product launch, accepted a lower bid from a buyer more likely to preserve the charter purpose, or allocated value to the public-benefit program, so long as the decision was informed, disinterested, and not waste. Anthropic’s own draft and public disclosures say the quiet part: as a PBC it may make decisions that appropriately balance the three interests and still hurt short-, medium-, or long-term performance and the Class A price, and it cannot guarantee a conflict will be resolved in stockholders’ favor. [17] [18] [3] Stockholder pecuniary interests remain one of the three weights. They are not the tie-breaker.

The biennial benefit statement under DGCL § 366

Section 366 is a disclosure rule, not an audit rule. A PBC must include in every notice of a stockholder meeting a statement that it is a public benefit corporation. No less than biennially, it must provide stockholders a statement on its promotion of the chartered benefit and of the best interests of those materially affected by its conduct. That statement must include four things: the objectives the board has set; the standards the board has adopted to measure progress; objective factual information based on those standards; and the board’s assessment of success. [5]

A third-party standard or certification, more frequent reporting, and public availability are optional. They apply only if the certificate or bylaws require them. [5] Delaware deliberately did not import the B Lab model. A PBC is a legal form; Certified B Corporation status is a private certification, and Anthropic’s charter does not appear to require an outside standard. [11]

Anthropic has chosen to do more than the floor, on its own terms. It published a 2026 Public Benefit Report and said it will publish annually, measuring itself against the five pillars. The report is a company narrative of progress toward the charter, not a third-party assurance report. [7] [8] Secondary readings of the S-1 report that the charter and bylaws will not require an objective third-party standard and that the board expects the standards to be qualitative, with the biennial statement made available on the website. [19] [18]

What this means for public stockholders. Once listed, holders are entitled to a benefit statement at least every two years, and Anthropic has said it will report more often and publicly. They are not entitled to an independent score of whether “the long-term benefit of humanity” was achieved. The standards are set by the same board whose conduct the statement describes. The report is evidence a plaintiff might later use; it is not itself a remedy.

How Delaware limits suits to enforce the benefit purpose

Section 367 is the enforcement gate, and the 2020 amendments made it wider in coverage and no easier to pass.

Current text, 8 Del. C. § 367:

Any action to enforce the balancing requirement of § 365(a) of this title, including any individual, derivative or any other type of action, may not be brought unless the plaintiffs in such action own individually or collectively, as of the date of instituting such action, at least 2% of the corporation’s outstanding shares or, in the case of a corporation with shares listed on a national securities exchange, the lesser of such percentage or shares of the corporation with a market value of at least $2,000,000 as of the date the action is instituted. This section shall not relieve the plaintiffs from complying with any other conditions applicable to filing a derivative action including § 327 of this title and any rules of the court in which the action is filed. [5]

Three features of that sentence matter more than the headline numbers.

The threshold is the lesser of 2% or $2 million, not “2% or $2 million” as alternatives a plaintiff must clear. For a corporation listed on a national securities exchange, plaintiffs must own the lesser of 2% of outstanding shares or shares with a market value of at least $2,000,000, measured on the date the action is instituted. For a company of the scale Anthropic is seeking to list, 2% is worth vastly more than $2 million, so the binding number is $2 million of market value. For a small listed PBC where 2% is worth less than $2 million, 2% is the binding number. Unlisted companies face a flat 2% test, with no dollar alternative. [5] [20]

The gate covers every theory that is really a balancing claim. The original 2013 version of § 367 limited derivative suits. The 2020 amendments (82 Del. Laws, c. 256, § 19, effective July 16, 2020) rewrote the section so that any action to enforce the § 365(a) balancing requirement—individual, derivative, or any other type—requires the same ownership. A plaintiff cannot restyle a balancing complaint as a direct claim and skip the threshold. The 2020 changes also clarified, in § 365(c), that a balancing failure is not bad faith or a loyalty breach for exculpation purposes unless the charter says so. [21] [22]

Ordinary derivative conditions still apply. Section 367 does not displace DGCL § 327 (the plaintiff must have been a stockholder at the time of the challenged transaction, or received the shares by operation of law from someone who was) or Court of Chancery Rule 23.1 demand requirements. Meeting the dollar or percentage test is necessary, not sufficient. [5]

Who is outside the gate entirely:

• Non-stockholders. Section 365(b) creates no duty to them, and § 367 confers standing only on stockholders who meet the ownership test. Mission beneficiaries, NGOs, employees as such, and customers cannot sue to enforce the purpose. [5] [11] [22]
• A retail holder, or a group of them, below the threshold. For a listed company the practical cutoff is $2 million of stock on the filing date, held individually or collectively.
• Claims that are truly about the balance, even if pleaded as something else. The “including any individual, derivative or any other type of action” clause is aimed at that evasion.

Who is inside, and what they still have to prove:

• Any holder or group with $2 million of listed stock (or 2%, if that is less) can institute an action to enforce the balancing requirement, subject to contemporaneous ownership and demand rules.
• They then have to get past § 365(b). Drakes Landing put the pleading burden on the plaintiff to allege facts supporting a reasonable inference that the decision was uninformed, interested, or one no person of ordinary, sound judgment would approve. A complaint that attacks only the pecuniary side of the balance, and says nothing about whether the board informed itself about affected constituencies and the chartered benefit, fails. [10]
• Classic fiduciary claims that are not actions “to enforce the balancing requirement”—for example, a self-dealing transaction in which a director took a personal benefit—are not gated by § 367. But § 365(c) blocks the most obvious workaround: absent a real conflict, a balancing failure cannot be recharacterized as non-exculpated bad faith or disloyalty. [5]

Commentary treats this design as intentional. The 2020 amendments were described at the time as “more broadly limiting stockholder litigation,” confirming that non-stockholder constituencies cannot sue and that only stockholders above the ownership floor can challenge the balance. [22] Practitioner guides state the same rule in operational terms: only stockholders can enforce the balancing duty, and only above 2% or, for a listed company, the lesser of 2% or $2 million; the public and the mission’s beneficiaries cannot. [11] Gibson Dunn’s note on Drakes Landing summarizes § 367 the same way: a 2%-of-shares threshold, or for listed companies the lesser of 2% or $2 million in market value. [10]

There is still almost no case law applying § 367 itself. Drakes Landing is the first substantial Chancery opinion on PBC fiduciary duties, and it was decided on the safe harbor, not on standing. Anthropic’s S-1, as summarized by commentators reading the filing, flags the flip side as a risk: § 367 opens a category of balancing suits that does not exist for a traditional corporation, and there is limited precedent on how a court would review the balance. [19] [10]

What this means in practice for Anthropic’s public stockholders.

The enforcement regime and Anthropic’s capital structure point the same direction. Litigation is a weak check. Governance is the real one, and public holders do not hold it.

• A Class A holder who wants to sue because the board subordinated the stock price to the charter purpose needs, with co-plaintiffs, at least $2 million of stock on the day of filing, plus contemporaneous ownership, plus a demand or demand-futility pleading, plus facts that defeat an informed-and-disinterested safe harbor that Drakes Landing applied at the motion-to-dismiss stage. That is a real path for an institution. It is not a path for a retail holder, a safety advocacy group, or “the public.”
• Even a successful balancing suit would ask a court to second-guess a three-way balance under a standard commentators describe as nearly contentless. Hamermesh’s “good luck” is the enforcement problem in one phrase: a court has no metric for whether “long-term benefit of humanity” was given enough weight relative to the stock price. [16]
• The people the charter purports to serve have no standing at all. If public holders believe the board is ignoring the purpose in favor of growth, qualifying stockholders can try to sue. If the public believes the board is ignoring the purpose, Delaware gives them the biennial (or, in Anthropic’s case, annual) report and nothing else.
• At Anthropic the practical substitute for stockholder enforcement is the Long-Term Benefit Trust. The S-1 describes the LTBT as a financially disinterested body that, through a single share of Class T common stock with no economic rights and no general voting rights, elects and removes the Class T directors, who constitute a majority of the authorized board. A separate Class F share, held by the founders’ LLC, carries 50.1% of the voting power on matters submitted to stockholders. The filing says Class A holders generally will not be able to affect the outcome of stockholder votes or director elections. [2] [18] Anthropic has said this openly: PBC status gives directors legal latitude to balance, but does not by itself make them accountable to other stakeholders, which is why the trust exists. [9] [23]
• PBC status itself is not locked. Since the 2020 amendments, conversion into or out of PBC status no longer requires a supermajority, and the old appraisal right for objecting stockholders was eliminated. [10] [21] At Anthropic that statutory flexibility is academic for public holders. Charter amendments and director elections run through a board majority elected by the trust and a founder vote of 50.1%.
• The economic commitments sitting beside the legal duty—the Anthropic Public Benefit Fund’s 846,840,960 Class N shares, described in the S-1 as 5% of fully diluted shares as of June 30, 2026, plus a discretionary framework for further issuances—are allocations the board controls, not a stockholder right to compel mission spending. [2]

The structure public stockholders are being asked to accept is therefore coherent, and it should be read as such. Delaware authorizes the board to trade their pecuniary interests against a charter purpose and against affected stakeholders, protects directors who do that trade in an informed and disinterested way, bars the beneficiaries of the purpose from suing, and limits stockholder suits to holders of $2 million (the lesser figure, once listed) who can also clear the safe harbor. Anthropic then places board control in a trust and voting control in the founders, so that the people most likely to use even that limited suit are not the people who pick the directors. The biennial statement tells stockholders how the board says the balance is going. It does not let them redo it.

Anthropic’s Long-Term Benefit Trust (LTBT) is a Delaware noncharitable purpose trust that holds the company’s only Class T share. That share has essentially no economic rights. Its job is to elect and remove a board majority, while a separate founder vehicle controls most other shareholder votes. The structure has already been rewritten once for the IPO, and the override path is harder than the 2023 blog post implied—but it is not trustee-proof.

What the trust is, and what Class T actually is

The LTBT is not a stockholder in the ordinary sense. Under the Amended and Restated Trust Agreement filed with the S-1, it is a Delaware noncharitable purpose trust under 12 Del. C. § 3556. It has no beneficiaries. Its only purpose is to use the Class T share and related contractual rights so that Anthropic “responsibly balances” stockholder pecuniary interests, people materially affected by the company’s conduct, and “the responsible development and maintenance of advanced AI for the long-term benefit of humanity”—the same public-benefit purpose in the certificate of incorporation. [1] [2]

Anthropic created an earlier “Long-Term Benefit Committee” in its 2021 Series A documents, then replaced it. The current trust agreement is effective June 29, 2023. Class T was added to the charter at the Series C close. Anthropic described that publicly on September 19, 2023: five financially disinterested trustees, a growing right to elect and remove directors, and a majority of the board “within 4 years,” plus notice rights on actions that could significantly alter the company. [3]

Class T’s economics are deliberately empty. The S-1 capital-structure table says Class T has no dividend rights, is entitled only to par value in liquidation, does not vote on matters submitted to stockholders generally, and can be transferred only with approval of the “Trust Amendment Threshold.” It is also redeemable by the company for no consideration on specified events. A September 2026 charter amendment reverse-split outstanding Class T stock 1,000-for-1 into a single share; the post-IPO certificate authorizes exactly one Class T share. [2] [4] [5] The September 17, 2026 date for that reverse split is reported from the S-1 by secondary coverage; the amendment exhibit itself describes the split but the date was not independently extracted here.

The trust may not hold any other Anthropic securities. An institutional Administrative Trustee—The Bryn Mawr Trust Company of Delaware, on the signature page of the filed trust agreement—handles administration. Voting power sits with the Voting Trustees, who must be natural persons. [1] [6]

Who the trustees are, and how they turn over

Five Voting Trustee seats are authorized. Terms run until the next annual stockholder meeting, with no term limit; incumbents are reappointed by a majority of the other trustees then in office. New trustees are appointed by a majority of trustees then in office, after at least 30 days’ notice to the board’s Nominating Committee (unless waived). If the voting-trustee office is empty, 80% of the Whole Board can appoint a trustee. [1]

Removal is internal, with a narrow court backstop. A trustee can be removed with or without cause by a majority of the five authorized seats, or by unanimous approval of the other trustees then in office plus Nominating Committee approval, in each case after 30 days’ notice to that committee. The Delaware Court of Chancery can remove a trustee only on petition by an “Enforcer,” and only for a defined list (felony or dishonesty offenses, incapacity, addictive substance abuse that interferes with duties, willful misconduct, fraud, gross abuse of office, or a material confidentiality breach). Enforcers are a Voting Trustee, the company, the Class F holder, or a person or group that has held at least 10% of non-trust voting power for a continuous year. The public, as such, cannot enforce the trust. [1]

Conflicts are restricted but not absolute. A material pecuniary interest in Anthropic, a material competitor, a material counterparty, or an adverse party must be disclosed and approved by both the Nominating Committee and a majority of non-conflicted trustees. Trustee pay is cash only, fixed annually by the trustees, and may not vary with company profits or with any company action. [1] Secondary accounts of the S-1 say each trustee is paid $400,000 a year, with an extra $25,000 for the chair, and that Anthropic paid the trust $2.72 million in 2023, nothing in 2024, $906,089 in 2025, and $8,425,213 in the first nine months of 2026. Those dollar figures are widely reported from the S-1; they were not independently extracted from the S-1 HTML retrieved for this review, so treat them as reported, not re-verified here. [7] [8]

Roster, from Anthropic’s own announcements plus later reporting:

• Initial five, announced September 19, 2023, chosen by the board after a year-long search: Neil “Buddy” Shah (chair; Clinton Health Access Initiative), Jason Matheny (then RAND CEO), Kanika Bahl (Evidence Action), Paul Christiano (Alignment Research Center), Zach Robinson (then interim CEO of Effective Ventures US). Future trustees were to be elected by the trustees. [3]
• A mirrored capture of that same Anthropic page includes a footnote that Matheny stepped down in December 2023 over potential RAND conflicts and Christiano stepped down in April 2024 to lead AI safety at the U.S. AI Safety Institute. That footnote did not appear in the live-page extract retrieved here, so the dates are confirmed only via that capture and later secondary accounts, not by a fresh read of the live post.
• Richard Fontaine (Center for a New American Security) was announced June 7, 2025. Anthropic said new trustees are selected by existing trustees “in consultation with Anthropic” and hold no financial stake. Secondary S-1 summaries date his start to April 2025; the public announcement is June 2025. [9]
• January 21, 2026: Mariano-Florentino Cuéllar appointed; Bahl and Robinson “concluded their terms.” Shah credited them with helping appoint directors including Jay Kreps and Reed Hastings. [10]
• July 9, 2026 announcement: Ben Bernanke joined, alongside Shah, Fontaine, and Cuéllar. Anthropic again said trustees hold no equity, do not share profits, and are compensated only for time and service, and that new trustees are selected by existing ones in consultation with the company. Secondary S-1 summaries date his start to June 2026. [11]
• August 4, 2026: Cuéllar left the trust to become Anthropic’s first chief global affairs officer, reporting to Daniela Amodei. Reuters reported the trust would seek a successor. [12]
• October 21, 2026: Harvard physicist Christopher Stubbs appointed. CNBC listed the trust as Stubbs, Bernanke, Shah, and Fontaine. [13]

As of the October 26, 2026 S-1, secondary accounts that quote the filing say four of five seats are filled—Shah (chair, since June 2023), Fontaine, Bernanke, and Stubbs—with one vacant, and that “in practice” trustees and executives have collaborated on sourcing candidates. That vacancy and the “in practice” sentence are reported from the S-1, not re-extracted from the HTML text retrieved here. [8] [14]

The composition has shifted away from the 2023 AI-safety and effective-altruism cluster (Christiano, Robinson, Bahl, Matheny) toward public-health, national-security, macroeconomic, and scientific generalists. Only Shah remains from the founding slate.

How board power phases in, and what changed

The 2023 public description was a phase-in: Class T would elect a rising number of directors on time- and funding-based milestones, and a majority within four years. A Series C investor seat was added at the same time so later investors would keep a direct board voice. [3]

The pre-IPO restated certificate filed as an S-1 exhibit makes the schedule concrete:

• Before the First Phase-In Date (September 6, 2023): four directors, none elected by Class T.
• First Phase-In until the Second: five directors; Class T elects one.
• Second Phase-In until the Final: five directors; Class T elects two. Second Phase-In is the earlier of May 24, 2026, or four months after the board determines the company has raised $5 billion in “Total Funds Raised.”
• Final Phase-In until February 11, 2026: five directors; Class T elects three. Final Phase-In is the earlier of May 24, 2027, or eight months after $6 billion raised.
• From February 11, 2026: the authorized board is fixed at seven. Class T elects four; voting preferred elects one; voting common elects two. [15]

That February 2026 rewrite is the main structural change versus the 2023 design. Class T’s share went from a phased 1-of-5, then 2-of-5, then 3-of-5, to a hard 4-of-7, and the old “within four years / by May 2027” calendar majority was overtaken by a larger board in which Class T already holds four seats. Anthropic’s own April 14, 2026 announcement said that with Vas Narasimhan’s trust appointment, trust-appointed directors were a majority of the board, then listing seven directors: Dario Amodei, Daniela Amodei, Yasmin Razavi, Jay Kreps, Reed Hastings, Chris Liddell, and Narasimhan. [16]

The IPO charter, filed as Exhibit 3.3, replaces the milestone table with a standing rule: while Class T is outstanding, its holder elects “the smallest number of directors that constitutes a majority” of the authorized board, and the authorized number must be an odd number. The balance—“At-Large Directors”—is elected by Class A and Class F voting together. Class F, a single share to be issued to Founder LLC (the seven co-founders) immediately before the IPO, carries 50.1% of voting power on matters submitted to stockholders generally, including election of the non-Class T directors, and has no economic rights. Together, the trust and Founder LLC can elect the entire board. [5] [2]

The S-1 states that Class T directors “constitute a majority of our board of directors,” that the LTBT does not run day-to-day strategy, and that because the trust can elect a board majority Anthropic is a Nasdaq “controlled company.” It expects to use that exemption only for the nominating and corporate governance committee, which will not be fully independent. [2]

Post-offering seat math reported from the S-1, and consistent with the charter’s “smallest majority” rule on a seven-seat board: four Class T seats and three At-Large seats. Secondary accounts name the initial Class T directors as Reed Hastings, Vas Narasimhan, and Christopher Liddell, with one Class T vacancy, and the At-Large directors as Dario Amodei, Daniela Amodei, and Yasmin Razavi. TechCrunch’s October 27, 2026 account of the filing matches that board list and notes one board vacancy; it does not include Jay Kreps, who was on the April 2026 Anthropic list. Kreps’s departure is inferred from those two public lists, not from an Anthropic announcement retrieved here. [8] [14] [16]

After the Class F sunset—when no Class F share remains, or 450 days after Founder LLC has had fewer than three “Qualifying Members” for 90 consecutive days—the board splits into three staggered classes with three-year terms, allocated so each class has a proportionate mix of Class T and At-Large directors. Reuters reported that design as Anthropic’s plan for the period after the founders’ supervoting share is gone. [5] [17]

The one-seat brake, and the override

The majority is real, but one of the four Class T seats is not a clean trustee appointment. The post-IPO certificate designates one Class T director as the “Specified Class T Director.” A stockholders’-agreement exhibit filed with the S-1 designates Chris Liddell in that seat. Nominating him, removing him without cause, or changing who holds that designation requires either prior approval of the Nominating Committee or, if the committee does not approve within 15 days, a “Requisite Class E Vote.” The committee is three directors: one Class T director and two At-Large directors. Class E is a post-IPO employee class, excluding co-founders, with no economic rights and a vote only on specified matters. After the Class F sunset, the special consent rules for that seat fall away. Removal for cause is available as provided by law and is not subject to this brake. [5] [18]

The Wall Street Journal’s reading of the filing is that this one seat is the trust’s entire margin on a 4–3 board, and that the initial nominating committee is Daniela Amodei, Reed Hastings, and Yasmin Razavi. That committee composition is reported, not confirmed in a primary extract retrieved here. The mechanism itself is in the certificate. [19]

The 2023 post said most changes would be made by agreement of the trustees and the board, or the trustees and other stockholders, but that “failsafe” supermajorities could change the trust and its powers without trustee consent, with the required supermajority rising as the trust’s power phased in. [3] The IPO documents replace that vague failsafe with a defined Trust Amendment Threshold. Any of the following works, each required class voting separately: [5]

1. At any time: 80% of the Whole Board and the Class T holder (the trust itself).
2. Until the earlier of the Class F sunset or the third anniversary of the IPO close: the Class F holder, plus a Class E vote (majority of votes cast, 40% quorum), plus a majority of Class A votes cast (majority-of-outstanding quorum).
3. Until the Class F sunset: the Class F holder, plus a higher Class E vote (at least 75% of votes cast, majority quorum), plus the same Class A majority-of-votes-cast.
4. After the Class F sunset: 85% of outstanding Class A shares.

Class T can be transferred only with that threshold’s approval. A Voting Request approved by persons representing the Trust Amendment Threshold can instruct the trust how to vote or exercise Class T rights, including amending the trust agreement—but it may not designate specific individuals who must be elected as Class T directors. Before the Class F sunset, trustees also may not amend the trust agreement unless the Trust Amendment Threshold approves. [18]

The Wall Street Journal summarized the no-consent path as, for three years, unanimous founder approval plus a majority of employee shares voted plus a majority of Class A shares voted; then two-thirds of founders, three-quarters of employee shares voted, and a majority of Class A; and, once founder special powers are gone, 85% of Class A outstanding. The 85% figure and the employee-vote structure match the certificate. “Unanimous” and “two-thirds of founders” do not appear in the certificate text retrieved here, which requires the Class F holder as such. Those founder-internal percentages are reported, and may reflect the Founder LLC agreement, which was not reviewed. [20]

If Class T disappears, a “Class T Sunset Time” occurs. Sitting Class T directors’ terms expire immediately unless the At-Large directors have already approved otherwise, and thereafter all directors are elected by the Electing Holders—Class A and Class F together while Class F exists. The S-1, as reported by Reuters, says the point of these contingencies is to “preserve long-term flexibility if circumstances change,” and that directors would still be bound by the public-benefit provisions even if Class T passed to someone not bound by the mission. [5] [17]

Inside the trust, ordinary use of the Class T vote requires a majority of voting trustees then in office, and cannot be exercised if fewer than three voting trustees have been in office for more than 90 days. “Specified Matters”—certain charter and stockholders’-agreement provisions—require 80% of voting trustees then in office, or a Voting Request. [1]

What this means in practice

The trust’s power is board composition, not management. It does not set strategy, fire the CEO directly, or own the economics. It picks (and can remove) a majority of directors, receives contractual information and advice rights, and, under Anthropic’s Responsible Scaling Policy updates, has been given a formal role in external review of risk reports. The RSP change is a policy grant, not a charter power. [2] [21]

Three design choices matter more than the “independent majority” headline. First, one of the four seats can be blocked by a committee that includes two non-trust directors, with employees as the tie-break, until the founders’ Class F share sunsets. Second, founders still elect the other three directors and hold 50.1% of the general stockholder vote, so ordinary shareholders do not choose the board while both vehicles exist. Third, the trust can be instructed, amended, or stripped without trustee consent, but only through stacked supermajorities that get harder after year three and harder still after Class F is gone. The 2023 experiment was built to be difficult to unwind and possible to correct. The IPO documents keep both features, and make the correction path explicit.

---

Recent Findings Supplement (November 2026)

Anthropic filed its S-1 registration statement on October 26, 2026 (with an amendment on November 2), providing the most detailed primary disclosure to date on the Long-Term Benefit Trust (LTBT) structure ahead of its planned Nasdaq IPO under ticker ANT.[1][2]

This filing, along with contemporaneous trustee appointments and related exhibits (including the Amended and Restated Trust Agreement), constitutes the key recent developments since May 2026. Earlier details (e.g., general removal thresholds around 85% shareholder support) were largely unconfirmed reporting; the S-1 supplies precise mechanics.[3][2]

Trustees and Recent Appointments (2026)

As of the October 2026 S-1, the LTBT has four named trustees out of five authorized seats (one vacant):
- Chair: Dr. Neil “Buddy” Shah (CEO of Clinton Health Access Initiative; chair since June 2023).
- Richard Fontaine (joined April 2025; national security expert).
- Ben Bernanke (former Federal Reserve Chair; joined June 2026).
- Christopher Stubbs (Harvard experimental physicist and former dean of science; joined October 2026, first physicist on the trust; announced October 21, 2026).[1][4]

New data: Anthropic pays each trustee $400,000 annually in cash (fixed by the trustees, independent of company performance or profits); the chair receives an additional $25,000. The company funded the trust with $8.43 million in the nine months ended September 30, 2026 (vs. $906,089 for all of 2025), covering operations and compensation. Trustees hold no Anthropic equity or economic interest.[1][2]

Trustees are selected by majority vote of those in office (in consultation with the company in practice) and can be removed by majority of all authorized trustees or by unanimous vote of the others plus nominating/governance committee approval after notice. The trust is a Delaware noncharitable purpose trust with an administrative trustee (Bryn Mawr Trust Company of Delaware).[2][5]

Class T Stock, Board Election Power, and Structure

The LTBT holds the single outstanding share of Class T common stock (post a 1-for-1,000 reverse split effective September 17, 2026, as part of charter amendments). Class T carries no economic rights (no dividends, liquidation preference, or general voting) but generally grants exclusive authority to elect and remove the Class T directors, who constitute a majority (4 of 7) of the post-IPO board.[6][5]

Initial Class T directors (per S-1): Reed Hastings, Vas Narasimhan, Christopher Liddell, and one vacancy. The remaining three (At-Large) directors are elected by Class A and Class F shares voting together (initially including Dario Amodei, Daniela Amodei, and Yasmin Razavi). The LTBT also receives advance notice of certain major actions and advises on AI risks/societal impacts but has no day-to-day management role.[5][7]

Anthropic qualifies as a “controlled company” under Nasdaq rules due to the LTBT’s board majority power and plans to rely on the exemption only for its nominating/governance committee.[2]

Amendment, Override, and Sunset Provisions (Newly Detailed in S-1)

The S-1 outlines the “Trust Amendment Threshold” for changing LTBT terms, Class T rights, or related charter provisions—far more specific than prior reports:
- If trustees support the change: Requires 80% of the authorized board and 80% of trustees in office.
- If trustees oppose (while Class F exists): First three years post-IPO requires unanimous approval of continuing founders (via Class F), plus majority of Class E (employee) votes cast (with 40% quorum), plus majority of Class A votes cast. Later: 66⅔% of continuing founders, 75% of Class E votes cast, plus majority of Class A.
- After Class F sunsets (10 years post-IPO or if fewer than three continuing founders remain for 90 days): 85% of outstanding Class A shares.[3][2]

Class T itself can only be transferred with amendment-threshold approval. The structure includes employee (Class E) voting rights on certain LTBT-related nominations (e.g., one “Specified Class T Director” requires nominating committee concurrence or employee vote) and other checks.[8]

Note on provenance: Core structural elements (Class T mechanics, board split, purpose, trustee pay/funding, amendment thresholds, reverse split) are confirmed in the S-1 and Trust Agreement exhibit. Pre-filing reports on exact override percentages or trustee mechanics were secondary and are now superseded or refined by primary documents. No major changes to the trust’s core design were announced between May and October 2026 beyond trustee additions and the S-1 disclosures.[9]

These filings represent the primary recent updates; the trust remains independent of management/investors while embedding long-term mission oversight alongside founder voting control via Class F (50.1% aggregate vote).

Anthropic’s public S-1 (filed October 26, 2026, amended November 2, 2026) is the controlling map. It does not sell control with the economics. After the IPO, five common classes split money from power, and every outstanding preferred series is designed to disappear into Class A or Class N at the offering. Public buyers of Class A get one vote and a pro rata claim on dividends and liquidation—and, while Class F exists and the founders agree how to cast it, they do not get the outcome of ordinary stockholder votes or of most director elections.

The filing is public: Form S-1, SEC file 333-299633, CIK 0002130357, October 26, 2026, with an S-1/A on November 2, 2026. Price range and offering share count were still blank in the October filing. Class A is proposed for Nasdaq Global Select Market under “ANT.” The company is a Delaware public benefit corporation. It will be a Nasdaq “controlled company” because the Long-Term Benefit Trust, as sole holder of Class T, generally elects a board majority—not because the founders’ Class F share exists. Anthropic says it expects to use that exemption only for the nominating and corporate governance committee; audit and compensation committees are described as fully independent.

Post-offering common stock: five classes, two of which carry the economics

Immediately before the offering closes, the charter that becomes effective authorizes 460,000,000,002 shares at $0.00001 par: 300 billion Class A, 100 billion Class N, one Class T, one Class F, 30 billion Class E, and 30 billion undesignated preferred. Only Class A and Class N participate in dividends and in liquidation proceeds (pro rata with each other, after debt and any future preferred preferences). Class T, Class F, and Class E get par value in a liquidation and no dividends.

Class A — listed common. Held by public buyers, co-founders, employees (via RSUs and options), and other existing holders whose stock converts into voting common. One vote per share on matters submitted to stockholders generally, including election of the directors who are not Class T directors. Not convertible. Generally transferable, subject to lock-ups. Because Class F is entitled to a fixed 50.1% of total voting power on those general matters, all Class A together is capped at 49.9% of that voting power, and new Class A issuance does not dilute Class F. If a majority of Continuing Founders cannot agree how the Founder LLC should vote Class F, Class F is voted proportionately with the Class A votes actually cast—so a founder deadlock is the main path by which Class A can decide a general matter.

Class N — economic twin, non-voting. Same dividend and liquidation rights as Class A. No vote except where Delaware law requires a series vote (for example, a charter amendment that hits Class N differently from Class A). Holders are “primarily strategic business partners,” plus the Anthropic Public Benefit Fund (846,840,960 Class N shares contributed, equal to 5% of fully diluted shares as of June 30, 2026, with board authority to add up to another 2% each time the stock rises 50% above the prior milestone, starting from the IPO price) and donation-matching stock. Automatic 1-for-1 conversion into Class A on (i) a registered offering or open-market sale, or (ii) a transfer to a third party that is not a Class N holder or its affiliate and that will not beneficially own more than 9.99% of any class or series after the transfer. Not redeemable. Amazon and Google are expected, after preferred and note conversion, each to hold more than 5% of outstanding capital stock in Class N; exact counts were blank in the October filing.

Class T — one share, held by the Long-Term Benefit Trust. No economic rights. No vote on general stockholder matters. It elects and removes the Class T directors, who are a majority of the authorized board (four of seven). It also votes on redemption of Class E and on certain amendments to Class F rights. Transferable only with approval of the persons who make up the Trust Amendment Threshold. Not convertible. Redeemable for no consideration under the thresholds in the sunset section below. No calendar expiration.

Class F — one share, issued immediately before the offering to Founder LLC. No economic rights. On matters submitted to stockholders generally—including election of the non-Class T directors—it casts a number of votes equal to 50.1% of total voting power of outstanding capital stock on the record date. That percentage is fixed; it does not shrink as Class A or Class N is issued. Not convertible. Transferable only in a board-approved reorganization or redomiciliation of Founder LLC that does not change its members. Voted as the Founder LLC Agreement provides: generally a majority of Continuing Founders, each with one vote inside the LLC, separate from how each founder votes any personal Class A. Initial members expected to be the seven co-founders: Dario Amodei, Daniela Amodei, Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish, and Christopher Olah. Benjamin Mann is not a member and has no say in the Class F vote. As of September 30, 2026 (figures adjusted for the September 17, 2026 10-for-1 split of Class A and Class N), beneficial Class A counts were roughly 231.8 million to 245.3 million per co-founder; percentage columns were blank. Those Class A shares are voted separately from Class F. Each co-founder’s pledge to give away 80% of personal equity is described as voluntary and revocable, not a binding obligation. A Continuing Founder who retains less than 2% of IPO-date Class A can be removed from the LLC.

Class E — employee ballot stock, excluding co-founders. No dividends; par value only in liquidation. Not convertible and not transferable. No vote on general matters. One vote per share only on: (1) nomination, change, or removal of the Specified Class T Director when the nominating and governance committee does not concur with the LTBT, and (2) certain changes to the LTBT or Class T rights, and redemption of Class T. Issuance mechanics reported from the filing: one share per six months of service, plus shares tied to employee pledges under the donation-matching program; the company said it would decide timing after the offering. None were outstanding on the September 30, 2026 pro forma capitalization (which assumed Class F issuance and other conversions, with share counts otherwise blank). After a departure other than for cause, half of that employee’s Class E is redeemed annually for nothing; all of it is redeemed on death, incapacity, termination for cause, or joining a direct competitor. All Class E is redeemed for nothing when Class F is gone, or earlier with 80% of the board plus Class F plus Class T.

Board election is a different denominator from the 50.1% vote

Seven authorized directors. The LTBT’s Class T share generally elects four Class T directors. Reporting on the October filing names Reed Hastings, Vas Narasimhan, and Christopher Liddell (the Specified Class T Director), with one seat vacant. Liddell cannot be replaced without cause unless the nominating committee concurs or, if it does not within the stated window, Class E approves. The other three directors—Dario Amodei, Daniela Amodei, and Yasmin Razavi—are elected by Class A and Class F voting together. Because Class F holds 50.1% of general voting power, Founder LLC determines those three seats if the Continuing Founders agree. The LTBT cannot remove the Amodeis. Together, LTBT and Founder LLC can elect every director. The S-1 states that Class A holders generally will not be able to affect the outcome of any matter submitted to stockholders or the election of directors, regardless of how they vote, while both control classes exist.

Trustees are not the Class T directors. The LTBT is a Delaware noncharitable purpose trust with five authorized trustee seats; trustees may not hold Anthropic securities. Serving trustees named in coverage of the filing include chair Neil “Buddy” Shah, Richard Fontaine, Ben Bernanke, and Christopher Stubbs, with one vacancy. The company funds the trust.

How the control shares end

Neither Class T nor Class F has a fixed calendar sunset.

Class F sunset. Redeemable for nothing (i) at Founder LLC’s option on supermajority approval of Continuing Founders, (ii) on dissolution of Founder LLC, or (iii) at the end of a Class F Transition Period. That period starts when fewer than three Continuing Founders (excluding certain “Non-Qualifying” successor members) have remained for 90 consecutive days, and runs 450 days. It is not extendable or waivable. After Class F is gone, if Class T remains, LTBT still elects the Class T majority; Class A elects the rest and decides general matters. Class E disappears with Class F. The filing contemplates that the board then moves to three staggered classes with three-year terms, and that the company opts into DGCL §203.

Class T redemption / “Trust Amendment Threshold.” For no consideration:
- If trustees vote in favor: 80% of authorized directors and 80% of trustees then in office (0.5 rounds up).
- If Class F is still out and trustees oppose: until the third anniversary of the offering, unanimous Continuing Founders (via Class F) plus a majority of Class E votes cast (40% quorum) plus a majority of Class A votes cast (majority quorum). At any time, 66⅔% of Continuing Founders plus 75% of Class E votes cast (majority quorum) plus a majority of Class A votes cast (majority quorum).
- If Class F is gone and trustees oppose: 85% of outstanding Class A.

These are separate class consents, not a combined vote in which Class F’s 50.1% swamps Class A. That is why Class A can block a hostile unwind of the trust even though it cannot win an ordinary majority vote.

Preferred stock: many series now, none after the IPO

The October S-1 states that after the offering, no preferred will be outstanding. Voting preferred converts into Class A; non-voting preferred converts into Class N. Conversion is optional at any time and automatic on a qualified firm-commitment IPO. As of June 30, 2026 the conversion ratio was 10:1 for every series except Series D-1, D-2, and D-3 (about 10.6:1, 10.4:1, and 10.6:1). A 10-for-1 forward split of Class A and Class N took effect September 17, 2026 and is reflected retroactively in common-share figures; preferred was not split, which is why Series H still prints at $589.0095 pre-split, or about $58.90 per post-split Class A share. No dividends had been declared through June 30, 2026. The pre-IPO certificate provided a 6% non-cumulative preference on original issue price ahead of common. In a liquidation or deemed liquidation, preferred is entitled to original issue price ahead of common, with deemed conversion if converting would pay more—standard non-participating preferred, not a participating double dip.

Voting redeemable convertible preferred — vote with Class A, convert to Class A. Series A, C-1, D-1, D-2, E-1, F-1, G-1, H-1. Each share votes as the number of Class A shares it could convert into, and these series vote together with Class A as a single class. They have no separate right to elect the board beyond that as-converted vote.

Non-voting redeemable convertible preferred — convert to Class N. Series B, C-2, D-3, E-2, E-3, E-4, E-5, F-2, F-3, F-4, G-2, G-3, G-4, H-2, H-3, H-4. No vote and no right to vote in board elections, except as required by law or specifically set in the certificate.

Selected issuance facts from the notes (share counts in full shares where secondary write-ups quote the S-1 directly; the financial-statement table is in thousands and matches): Series H, May–June 2026, $589.0095 per share, about $65.2 billion gross—77,637,878 Series H-1 to 305 investors and 33,110,291 non-voting Series H-2 to 11 investors. Amazon put in $5.0 billion and Google $10.0 billion of that round. Earlier voting-round prices in the notes: Series E-1 about $56.09 (January–March 2025), Series F-1 about $140.97 (August–September 2025), Series G-1 about $259.14 (January–February 2026).

Notes that become Class N, not a new class. Amazon 2023 and 2024 note purchase agreements convert at the IPO into Class N at the lower of a fixed price ($2.0928 and $5.0478 post-split, respectively) and the IPO price, subject to a waivable ownership cap; amounts above the cap roll into non-interest-bearing notes and convert quarterly at the lowest of the fixed price, the IPO price, and a 10-day trailing VWAP. Secondary write-ups of the filing put remaining principal near $2.2 billion (2023 rollover) and $4.0 billion (2024 notes), implying up to about 1.84 billion Class N shares if fully converted at those fixed prices. Google’s older notes had already converted into non-voting preferred (Series E-3, E-4, E-5, G-4) at discounts to the round price. Undrawn facilities (Google 2026 NPA up to $30 billion, reduced to $20 billion after the Series H check; Amazon 2026 NPA up to $20 billion, reduced to $15 billion) can settle in non-voting stock or notes. Those are contractual conversion rights, not separate stock series.

Blank preferred after the IPO is a real protective provision—for the founders, not for public holders. The post-offering board (80% of authorized directors) can create series and set their rights without a stockholder vote, but until the Class F sunset it cannot designate or issue any preferred without Founder LLC’s affirmative vote or consent. If voting preferred were later issued, Class A’s share of general voting power would shrink and Class F would stay at 50.1%.

Investor protective provisions that attach to today’s preferred (special class votes on mergers, new senior stock, charter amendments, and similar) die with automatic conversion. They are not carried forward as rights of Class A or Class N. What replaces them is the post-IPO charter: Founder LLC consent to new preferred before the Class F sunset, DGCL class votes if a class’s rights are changed adversely and unequally, and the multi-party thresholds to redeem Class T or Class F.

What Class A / public common can and cannot vote on

While both Class F and Class T are outstanding—the structure the S-1 says will last “for the foreseeable future”:

Matter	Class A can	Class A cannot
Ordinary stockholder votes (charter amendments submitted generally, most M&A requiring a stockholder vote, say-on-pay style items, auditor-type votes)	Cast 1 vote per share, in aggregate no more than 49.9%	Determine the outcome if Continuing Founders agree on the Class F vote. The filing says holders generally cannot affect the result regardless of how they vote
Founder deadlock on Class F	Decide the matter, because Class F then follows the Class A votes cast	—
Election of 4 Class T directors (board majority)	Nothing	Elect or remove them. That power sits in one Class T share
Election of the other 3 directors	Vote with Class F as a single class	Win the election against a united Founder LLC
Sale of the company	Vote, but lose a simple-majority stockholder vote to Class F; a sale also needs board approval, and LTBT elects most of the board	Force or block a sale by stockholder vote alone
Dividends	Receive pro rata with Class N if the board declares them	Compel a dividend
Conversion of Class N (Amazon, Google, Public Benefit Fund)	Nothing until a qualifying transfer	Keep partner stock non-voting; a qualifying sale turns it into Class A and adds to the voting float, still under the 49.9% cap
Redeem or neuter Class T over trustee objection	Block it by withholding the required majority of Class A votes cast (with quorum). That consent is a separate class approval, not swallowed by the 50.1%	Do it alone. Needs founder and Class E thresholds as well, and for three years the founders must be unanimous
Redeem Class T with trustee consent	Nothing required of Class A	Stop an 80% board / 80% trustee redemption
New preferred with voting rights	Nothing, until Class F sunsets	Stop issuance. Board plus Founder LLC can create it, and Class F’s 50.1% would not dilute
Controlled-company exemptions	No stockholder vote restores majority-independent board or independent nominating committee while LTBT elects a majority	Rely on Nasdaq’s ordinary governance package. Company says it will use the exemption only for nominating/governance
PBC balancing duty (stockholders vs. affected persons vs. “responsible development and maintenance of advanced AI for the long-term benefit of humanity”)	Sue under DGCL §367 only if the holder has at least 2% of outstanding shares or $2 million of stock, whichever is less. Director balancing is upheld if informed, disinterested, and not irrational	Force the board to prefer pecuniary interests. The S-1 says a conflict is not guaranteed to be resolved in stockholders’ favor
After Class F sunset, Class T still out	Elect the non-Class T directors and decide general matters	Elect the board majority; remove the trust without 85% of outstanding Class A if trustees object
After both control shares are gone	Elect the full board and decide general matters, subject to a staggered board and DGCL §203 if adopted as described	—

The practical reading of the S-1 is that a Class A share is an economic claim plus a contingent governance claim. The contingency is a founder split, a Class F sunset, or a special consent on unwinding the trust. It is not a claim on who runs Anthropic the day after the IPO.

---

Recent Findings Supplement (November 2026)

Anthropic, PBC publicly filed its Form S-1 on October 26, 2026 (amended November 2, 2026), disclosing a novel five-class common stock structure and “controlled company” status ahead of its planned Nasdaq IPO under ticker “ANT.”[1][2]

This filing provides the first comprehensive public details on capitalization, voting, board rights, conversion, and protective provisions—superseding prior private disclosures or reporting. No IPO has occurred as of November 4, 2026; the document leaves share counts and price range blank while outlining the governance framework designed to separate economic ownership from control in service of its public-benefit mission.[3]

Anthropic maintains five classes of common stock post-IPO (authorized but with specific issuances and conversions from preferred stock and other instruments). Preferred stock (various Series A–H) automatically converts into Class A or Class N common stock immediately prior to the offering. Class T and Class F each consist of a single share issued pre-closing.[4]

• Class A common stock: One vote per share on all matters submitted to stockholders generally (including election of non-Class T directors). Full economic rights (pro rata dividends and liquidation proceeds). Held by public investors (the shares offered in the IPO), existing holders, and co-founders/employees. Not convertible. Class A holders collectively hold up to 49.9% of total voting power.[5]
• Class N common stock: No voting rights except as required by law. Full economic rights identical to Class A. Primarily held by strategic partners (e.g., Amazon, Google via convertible notes/preferred); generally converts to Class A upon sale/transfer to an unaffiliated third party. Used for certain public-benefit issuances (e.g., 5% of fully diluted shares contributed to the Anthropic Public Benefit Fund as of mid-2026, with potential additional 2% issuances tied to share-price milestones).[6]
• Class T common stock: One share held by the Long-Term Benefit Trust (LTBT; five financially disinterested trustees with AI safety/national security/public policy expertise). No general voting rights on stockholder matters. Authority to elect/remove Class T directors (a board majority). No economic rights (par value only on liquidation). Enables “controlled company” status under Nasdaq rules.[7]
• Class F common stock: One share held by Founder LLC (initial members: seven co-founders Dario Amodei, Daniela Amodei, Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish, Christopher Olah). Entitled to votes equal to 50.1% of total voting power on matters submitted to stockholders generally (including election of non-Class T directors). Fixed percentage (not diluted by new Class A issuances). No economic rights. Class F and Class A (together) elect the remaining three directors.[8]
• Class E common stock: Issued to employees (excluding co-founders). One vote per share but only on limited matters (e.g., certain amendments affecting LTBT/Class T/Class F, redemption events, nomination/removal of a specified Class T director). No general economic rights beyond par value on liquidation; subject to redemption/forfeiture upon cessation of service or specific triggers (e.g., board/Trust/Founder LLC approvals or founder attrition thresholds).[9]

Board composition and election rights flow from the dual-control structure (LTBT via Class T for majority; Founder LLC via Class F for the balance). The LTBT elects/removes Class T directors (majority of the board); Class F/Class A elect the rest (including specific named directors like the Amodeis and Yasmin Razavi in some descriptions). This setup, combined with PBC status (balancing stockholder interests with the public-benefit purpose of responsible advanced AI development), limits public influence.[5]

Conversion terms are primarily automatic pre-IPO for preferred stock (to Class A or N depending on series); Class N converts to Class A on unaffiliated transfer. No routine conversion rights for the special classes. Protective provisions exist for Class E (e.g., adverse amendments require specific approvals including Requisite Class E Vote) and generally for governance changes involving the Trust or Founder LLC (notice/approval thresholds). Class F has notice rights on certain Trust amendments.[9]

Anthropic qualifies as a Nasdaq “controlled company” due to the LTBT’s board-majority election power, allowing exemptions from certain independence requirements (it plans to rely only on the nominating/governance committee exemption and maintains a majority-independent board with independent audit/compensation committees).[7]

For public Class A (“public common”) holders post-IPO: They receive one vote per share on general matters and non-Class T director elections but hold only 49.9% aggregate voting power. They generally cannot determine outcomes of stockholder votes or director elections. They have no say in electing the Class T directors (board majority) or overriding the fixed Class F control. Influence is further limited on structural/governance matters tied to the PBC mission, LTBT, or Founder LLC. They hold standard economic rights but face warnings in the filing that the structure may limit ability to affect important decisions, including change-of-control scenarios.[5]

Summary table of Class A/public common holder rights (post-IPO, based on S-1 disclosures):

• Can vote on: General stockholder matters (one vote/share); election of non-Class T directors (as part of the 49.9% bloc).
• Cannot meaningfully control/affect: Election or removal of Class T directors (board majority); overall outcomes of most votes due to 49.9% cap and Class F’s 50.1%; changes to core governance structure without alignment from Class F/T holders; redemption or key amendments affecting special classes.

No other material updates to capitalization or governance appear in sources from after May 4, 2026, beyond the S-1 and its amendment. Earlier private valuations (e.g., $965 billion post-money in May 2026 Series H) and compute commitments are referenced but do not alter the stock-class mapping.[10]

This structure prioritizes mission continuity over traditional shareholder democracy, with implications for any public-market entry or competitive governance benchmarking in AI.

Amazon and Google are Anthropic’s largest outside economic owners and its largest cloud counterparties, but the October 26, 2026 S-1 walls them off from formal control: their stock is non-voting Class N, contractual ownership caps sit well below a control threshold, and neither has a board seat or observer right. The real leverage is commercial, not corporate.

Control is deliberately separated from capital

Anthropic’s post-IPO structure treats Amazon and Google as funders and suppliers, not governors. Immediately before the offering, the company issues one Class F share to a Founder LLC whose initial members are the seven co-founders. That single share carries a fixed 50.1% of the vote on matters submitted to stockholders generally, including election of directors other than the Class T directors, and it has no economic rights. Class A holders together are capped at 49.9%. A separate Class T share, held by the Long-Term Benefit Trust, elects a majority of the board. Class N—the class “primarily” held by strategic partners—has the same economics as Class A but no vote except where law requires one, and it converts into voting Class A only on a sale or transfer to an unaffiliated third party.

The S-1 states that after preferred stock and convertible notes convert, Amazon and Google will each hold greater than 5% of outstanding capital stock, in non-voting Class N, and that those holdings “represent a significant economic interest.” Exact share counts and percentages are blank in the filing. Because the Trust can elect a board majority, Anthropic qualifies as a Nasdaq “controlled company.” It says it expects to use that exemption only for the nominating and corporate governance committee.

What this means for anyone trying to buy influence: writing a check does not buy a vote, a director, or a veto. The only path from Class N to voting stock is a transfer out of the Amazon or Google corporate family.

Ownership stakes are large, capped, and still not fully disclosed

The hard numbers in the S-1 are dollars and conversion mechanics, not percentages. Public estimates of the stakes disagree because later rounds diluted earlier holdings and because conversion is gated by caps the holders can waive.

Amazon. TechCrunch and Bloomberg both put cumulative cash invested at about $18 billion: roughly $8 billion of convertible notes from 2023 through 2025, plus $5 billion in each of the Series G and Series H preferred rounds in 2026. As of June 30, 2026, $6.2 billion of note principal remained outstanding and could convert into as many as 1.84 billion Class N shares (conversion prices cited in secondary analyses at about $2.09 and $5.05). Conversion stops once Amazon would exceed a 12% ownership cap on a partially diluted basis, unless Amazon waives the cap. Excess stays as non-interest-bearing rollover notes that convert in later quarters only as room opens under the same limit. An April 2026 facility still allows up to $15 billion more of non-voting stock or notes, drawn against compute milestones. On Amazon’s own books at June 30, non-voting preferred was carried at $92.5 billion and the notes at an estimated fair value of $97.9 billion.

Bloomberg’s reading of the filing puts Amazon’s current economic stake at roughly 9% on an as-converted, fully diluted basis, with the notes theoretically worth another ~11%—but the 12% ceiling binds unless waived. PitchBook has said full conversion without the cap would imply roughly one-fifth. Those figures are calculations, not S-1 disclosures.

Google. Court filings reported by The New York Times in 2025 put Google at about 14% on a fully diluted basis, contractually capped near 15%, with no voting rights, board seats, or board-observer rights. The 2026 S-1 shows further investment: about $10 billion of Series G-2 in March 2026 and $10 billion of Series H in May 2026, against a March 31, 2026 note-purchase agreement for up to $30 billion more through December 31, 2030, tied to operational and revenue milestones. After the Series H check, about $20 billion of that facility remained. Bloomberg calculates that Google has invested about $23 billion in total and holds an economic stake of at least 6.6% on an as-converted, fully diluted basis—excluding a separate $300 million investment in 2022 that the S-1 does not detail. A PitchBook analyst told TechCrunch the stake is closer to 8% after dilution from the 14% court-filing figure. The 2026 agreement caps ownership at 14.99% unless Google waives it; excess is funded through a zero-interest rollover note. Google also has an option to buy up to $10 billion of Class N at the IPO price in a concurrent private placement.

UK Competition and Markets Authority materials from the earlier Google partnership review described the stake as non-voting, with no board representation, and found that consultation rights did not amount to material influence sufficient for a merger finding. That is the design: stay under control thresholds that would turn a strategic investment into an acquisition review.

For a competitor or a public buyer of ANT, the relevant risk is not a sudden Amazon or Google takeover vote. It is that both can still choose to waive their caps, and that Class N becomes ordinary float the moment either sells to an outsider—so the “permanent” strategic capital is also future supply.

The commercial contracts are the real governance

The S-1’s binding constraints on Anthropic are multi-year, largely non-cancellable compute minimums and a distribution channel that already moves nearly half of revenue.

• Anthropic has committed to spend at least $110 billion with Amazon from May 2026 through April 2036, and at least $114.5 billion with Google from April 2026 through July 2033. Both are owed regardless of actual usage; shortfalls are generally trued up. Neither deal is cancellable except for an uncured material breach by the provider. AWS is described as the primary training and cloud provider for mission-critical workloads and supplies Trainium capacity, including the Project Rainier cluster. Google supplies Cloud capacity and TPUs; Anthropic has also estimated another $125–150 billion of additional TPU capacity from 2028 to 2031, outside the minimum table.
• Distribution is the second loop. Confidential prospectus figures reported by Reuters, and the public S-1, show sales through Amazon and Google cloud marketplaces at 47% of 2025 revenue (about $2.16 billion) and about 48% of first-half 2026 revenue (Amazon, Google, and Microsoft marketplaces together in the later period). Reuters’ analysis of the confidential filing put distribution fees at roughly $351 million in 2025, about 16 cents on each marketplace dollar. Those platforms also collected a rising share of customer receivables—60% of $909 million outstanding at the end of 2025.
• Chip exposure runs through Google’s design even when the lessor is someone else. The largest single equipment commitment in the filing is about $161.2 billion of Broadcom leases on TPU-based hardware. Broadcom can be issued up to $42 billion of notes solely to finance that rent.

The mechanism is circular by construction. Amazon and Google put capital in, often convertible into more Class N. Anthropic spends a contracted minimum back to them for compute whether or not usage matches the forecast. They also resell Claude, keep a fee, and see pricing and commercial terms. Their own models—Gemini at Google, Nova and related offerings at Amazon—compete with Claude on the same platforms.

How the IPO filing describes the conflict—and what it leaves out of “related party”

The S-1 does not hide the tangle. It describes Amazon, Google, and Microsoft as “investors, customers, distribution partners and competitors” at once. It warns that reliance on a limited set of cloud providers and suppliers, some of which are also strategic investors and customers, “may create complex situations that could lead to conflicts of interest and adversely affect our ability to secure computing power,” and that the relationships “create potential conflicts of interest that are difficult to fully anticipate or mitigate.” Specific channels of harm named or closely paraphrased in coverage of the filing:

• The cloud companies have “visibility into our pricing and commercial terms,” which can shape how they allocate capacity and how hard they sell Claude versus their own models.
• They can prioritize capacity for their own AI products or slow delivery.
• As shareholders they have a financial stake in Anthropic’s valuation, which can cut against hard bargaining—or, because Anthropic depends on them for both capital and compute, give them practical influence the charter denies them.
• The ties have already drawn antitrust scrutiny and could draw more.

There is a disclosure asymmetry that matters for public investors. Analyses of the S-1 note that Anthropic defines related parties by holdings of voting stock. Amazon and Google hold non-voting Class N, so the two partners with the largest stakes and the largest contracts are not in the formal related-party financing tables. Their post-IPO share counts are blank. The economic conflict is discussed at length in risk factors; it is not presented as a classic related-party transaction schedule with named percentages.

Anthropic’s own answer, repeated across years, is that dual (now multi) sourcing is what preserves independence. In a May 2024 Bloomberg interview, CEO Dario Amodei argued that partnering with both Google and Amazon produces more independence than reliance on a single hyperscaler. A February 2025 court submission in the Google antitrust case, filed on Anthropic’s side of the record, states that Google’s investments “give it no role in Anthropic’s governance and no control over its operations,” that Google has “no voting rights, board seats, or even board observer rights,” and no exclusive product rights—and that buying compute from multiple partners is how Anthropic avoids dependence on any one supplier. Amazon, when it took its second $4 billion tranche to $8 billion in November 2024, was described by Anthropic and by CNBC as remaining a minority investor with no board seat. The UK CMA reached a similar conclusion on the earlier Google stake: non-voting, no board seat, not a relevant merger situation.

That legal position is accurate on corporate formalities. It does not answer the commercial question the S-1 itself flags. A supplier that can withhold or reprice the capacity Claude is trained and served on does not need a board seat to affect strategy. The minimum-spend clauses make the dependence durable: Anthropic pays the Google and Amazon floors even if it shifts workloads elsewhere, which raises the cost of actually using the multi-cloud option the independence argument depends on.

What a competitor or public shareholder should take from this

Formal control is locked to the Founder LLC and the Long-Term Benefit Trust for the foreseeable future. Amazon and Google cannot elect directors, cannot outvote the Class F share, and cannot convert their own Class N into votes without selling it away. Caps at 12% (Amazon) and 14.99% (Google), both waivable by the holder, were built to stay under control and merger-review lines while still allowing very large economic exposure—Amazon alone marked its position at roughly $190 billion at June 30, 2026.

The contestable ground is capacity, distribution, and information, not the proxy card. Nearly half of recent revenue already clears through partners who also field competing models and who are contractually owed more than $224 billion of minimum cloud spend between them, before Broadcom TPU leases and other suppliers. The S-1 treats that overlap as a risk to compute access and to alignment of incentives, while keeping the same parties outside the voting-stock definition of related parties. For anyone underwriting ANT, the governance story and the commercial story point in opposite directions: the charter says neither Amazon nor Google can control the company; the contracts say both can make independence expensive.

---

Recent Findings Supplement (November 2026)

Anthropic’s October 26, 2026 S-1 IPO filing (with a November 2 amendment) provides the first detailed public accounting of its ties to Amazon and Google, confirming large non-voting stakes, strict ownership caps, zero board or observer rights, and explicit statements that neither investor can control the company.[1][2]

This filing, released just weeks before expected IPO marketing (potentially the week of November 9), adds fresh quantitative data on investments, compute commitments, revenue routing, and risk factors that were previously estimated or confidential. It underscores Anthropic’s structure as a public benefit corporation designed to insulate mission-driven governance from investor influence.[3]

Ownership Stakes and Recent Investments

The S-1 details Amazon’s and Google’s equity positions as primarily non-voting preferred stock and convertible notes, with explicit post-conversion caps that prevent control. Amazon’s total investment reaches approximately $18 billion (including ~$8 billion in notes from 2023–2025 and $5 billion each in Series G and H preferred in Q2 2026), while Google committed up to $40 billion ($10 billion upfront in March 2026 Series G-2 plus performance-linked amounts).[4][5]

• Amazon holds $6.2 billion in outstanding convertible notes as of June 30, 2026 (valued at $97.9 billion by Amazon; its overall Anthropic holdings carried at $190.4 billion). These convert into up to ~1.84 billion non-voting Class N shares, subject to a 12% ownership cap of Anthropic’s partially diluted share count (waivable by Amazon). Excess notes remain outstanding and can convert later if room opens under the cap. Post-IPO estimates range from 15–21% or roughly one-fifth if fully converted before the cap binds.[4][6]
• Google’s stake stood at ~14% per 2025 court filings (capped at 15% under its agreement, no voting rights or board seat). It added $10 billion in March 2026 and another $10 billion in May 2026; estimates of its current stake post-dilution range from ~7–8% to 10–15%. It holds an option for up to $10 billion in Class N shares at the IPO price and can draw up to $20 billion more under a March 2026 note facility through 2030 (performance-tied). Ownership capped at 14.99% (waivable).[1][2]
• Both convert primarily into Class N common stock post-IPO, which carries identical economic rights to Class A but no voting rights (except as required by law) until sold to an unaffiliated third party.[7]

These figures represent new, filing-backed precision beyond prior estimates and highlight how the stakes function more as commercial alignments than governance tools.

For competitors or new entrants: The caps and non-voting structure show Anthropic successfully attracted massive strategic capital while retaining founder control, but they also mean Amazon and Google have skin in the game that could influence pricing or capacity negotiations in their favor as suppliers.

Governance Limits, Board Rights, and Control Statements

The S-1 confirms Amazon and Google have no board seats or observer rights. Anthropic explicitly structures governance to prevent external control, with the seven co-founders holding 50.1% voting power via a single Class F share (through Founder LLC, no economic rights) and a Long-Term Benefit Trust electing a board majority via Class T shares.[7][5]

• Ownership caps (Amazon 12%, Google 14.99%) and Class N non-voting status directly limit influence; the S-1 notes these holdings represent a “significant economic interest” but stresses the structure ensures founders and the trust steer the company.[2]
• Anthropic describes its Public Benefit Corporation setup and founder supervoting as protecting its mission; the filing states that relationships with investors like Amazon and Google do not confer control.[7]
• The trust (independent trustees with no financial stake in Anthropic) elects a majority of directors; founders retain influence over key matters even after potential sunset of certain structures.[3]

This adds concrete governance details absent from earlier reporting.

Implications: Entrants cannot rely on equity stakes alone for influence; Anthropic’s model demonstrates how AI labs can raise billions from rivals while embedding mission-protective governance that sidelines investor voting power.

Commercial Agreements: Cloud, Chips, and Distribution

The filing quantifies minimum multi-year compute commitments (often payable regardless of usage) and revenue concentration through the partners’ marketplaces.[8]

• Amazon: Minimum $110 billion AWS spend commitment through April 2036 (10-year horizon cited in some reports). Includes Trainium chips and Bedrock distribution.[4]
• Google: Minimum $114.5 billion Google Cloud spend through July 2033; additional TPU capacity deals. Sells Claude via Vertex AI.[8]
• Combined with Microsoft ($44.9 billion): At least $269 billion in Big Tech compute commitments. Additional Broadcom chip leases (~$161 billion) and others push total infrastructure obligations much higher.[9]
• Distribution: 47–48% of revenue routed through Amazon and Google cloud marketplaces (e.g., ~$2.16 billion or 47% in 2025; ~48% in H1 2026), with ~16–17% fees retained by the platforms. These partners also collect a growing share of receivables.[10]

Spending on compute/infrastructure reached nearly $9 billion in H1 2026 alone.[6]

For market participants: These take-or-pay-style deals lock in revenue for the hyperscalers while creating deep interdependence; new cloud or chip providers face high barriers as Anthropic diversifies but remains heavily committed.

Conflicts of Interest and Related-Party Risk in the IPO Filing

The S-1 devotes substantial space (dozens of pages of risk factors) to the “complex dynamics” of these relationships, describing Amazon, Google, and Microsoft as simultaneous “investors, customers, distribution partners, and competitors.”[11]

• Key risks flagged: Potential conflicts where partners could favor their own models (e.g., Gemini), gain visibility into Anthropic’s pricing/terms, curtail compute access, or trigger further antitrust scrutiny. Dependence on a limited number of suppliers “could adversely affect our access to compute.”[10][12]
• Related-party transactions are detailed at length, including the equity investments, notes, and commercial deals. The filing notes these create “potential conflicts of interest that are difficult to fully anticipate or mitigate.”[11]
• Broader context: Anthropic also highlights AI safety/existential risks and its governance structure’s limits on ordinary shareholder influence.

This level of explicit, quantified disclosure on frenemy dynamics is a notable addition from the filing.

Competitive takeaway: The risks section serves as a blueprint for how dependent AI labs must navigate supplier-investor-competitor overlaps; pure-play entrants may face fewer conflicts but also less access to capital and distribution.

Overall, the S-1 transforms prior estimates into binding, public commitments and risk disclosures just ahead of Anthropic’s planned listing, highlighting both the scale of Big Tech entanglement and the deliberate governance firewalls Anthropic has erected. No major updates appear between the initial filing and early November 2026 reports.

The filing is public. Anthropic, PBC filed Form S-1 on October 26, 2026 (file 333-299633; accession 0001628280-26-067168) and Amendment No. 1 on November 2, 2026. Class A is proposed for the Nasdaq Global Select Market as “ANT.” Share count and price were still blank in the documents reviewed. The governance architecture below is taken from the October 26 prospectus text (cover, “Risks Related to our Governance Structure,” and the anti-takeover discussion), cross-checked against specialist readings of the same filing. The amendment updated operating results; it is not described as rewriting the control design.

What the prospectus actually sells is residual economics with almost no residual control. Two single shares, neither with economic rights, sit above the public float: Class T lets a purpose trust elect a board majority, and Class F lets seven co-founders cast 50.1% of the vote on everything else. Public Benefit Corporation (PBC) status then legally requires the board to balance profit against a charter purpose that is not profit.

The control stack: two shares, five classes, one public float

The cover calls the result “a novel capital and corporate governance structure” with “no direct public company precedent.” Immediately before the offering, Anthropic will issue one Class F share to Founder LLC, whose initial sole members are Dario Amodei, Daniela Amodei, Tom Brown, Jack Clark, Jared Kaplan, Sam McCandlish, and Christopher Olah. Benjamin Mann is not a member. That one share is entitled to 50.1% of total voting power on matters submitted to stockholders generally, including election of directors other than Class T directors, and it has no economic rights. The percentage is fixed, so later Class A issuance does not dilute founder votes. Inside the LLC, continuing founders vote one person, one vote; if a majority cannot agree, Class F is voted proportionately with Class A votes cast.

Class T, one share held by the Long-Term Benefit Trust (LTBT), generally elects and removes the Class T directors, who will be a majority of the authorized board, and otherwise does not vote on general stockholder matters and has no economic rights. Class A (the IPO stock) has one vote and pro rata dividends and liquidation rights. Class N, held primarily by strategic partners including Amazon and Google, has the same economics and no vote except where Delaware law requires a class vote; it generally converts to Class A on a transfer to an unaffiliated holder at or below 9.99%. Class E, for employees excluding co-founders, votes only on specified protective matters and has no economics.

Authorized capital in the certificate that becomes effective immediately before the offering is 300 billion Class A, 100 billion Class N, 30 billion Class E, 30 billion preferred, and one share each of Class F and Class T.

The prospectus states that, while both control classes exist, the LTBT and Founder LLC “will collectively have the ability to elect all members” of the board, and Class A holders “generally will not be able to affect the outcome of any matter submitted to a vote of our stockholders or the election of our directors, regardless of how they vote,” unless the continuing founders deadlock on Class F.

What this means for a buyer or a competitor: ANT is an economic claim, not a governance claim. Index funds that exclude unequal-vote or non-voting structures may be forced to pass, which the risk factors flag as a possible valuation drag. A rival that wants public-market accountability cannot copy the economics without copying, or explicitly rejecting, this split.

Board, trust, and the one real check

Post-offering the board is authorized at seven. Four Class T seats are identified as Reed Hastings (lead independent director), Vas Narasimhan, Christopher Liddell (the “Specified Class T Director”), and one vacancy. The other three, elected by Class A and Class F voting together—and therefore by Founder LLC in practice—are Dario Amodei, Daniela Amodei (chair), and Yasmin Razavi, with the filing’s director slate reflecting one vacancy in the current board as well.

The LTBT is a Delaware noncharitable purpose trust with five authorized trustees who may not hold Anthropic securities. As of the S-1, trustees are Neil “Buddy” Shah (chair; trustee since June 2023), Richard Fontaine (April 2025), Ben Bernanke (June 2026), and Christopher Stubbs (October 2026), with one seat vacant. Trustees choose their own successors. The prospectus says each trustee is paid $400,000 a year in cash, plus $25,000 for the chair, not tied to profits or to any corporate action. The trust does not run the company day to day; the board does. A board safety and risk committee is described as including Narasimhan and Liddell.

The real internal check is narrow and aimed at one seat. To replace Liddell, the Specified Class T Director, without cause, the LTBT needs the nominating and corporate governance committee’s concurrence or a Class E employee vote. Daniela Amodei sits on that committee. The LTBT cannot remove the Amodeis. That is why the trust is a board-majority mechanism, not a management-removal mechanism.

Class F has no calendar sunset. It ends if Founder LLC redeems the share or dissolves, or after a 450-day transition that starts once fewer than three continuing founders have remained for 90 consecutive days. A co-founder who retains less than 2% of their IPO-date Class A holdings can be removed from the LLC. Forbes reports that the LLC agreement, filed as an exhibit, lets each founder file a confidential ranked list of up to ten successors, approved by five of the other six; the governance committee can demand reconsideration but cannot veto an approved list. The first eight names must meet employment or public-benefit-organization criteria; the ninth and tenth need not. That succession mechanic is an LLC contract, not a charter term, but it is how Class F voting power is designed to outlive the original seven.

Ending Class T is harder. Readings of the filing describe a “Trust Amendment Threshold”: a consensual path (on the order of 80% of the board and 80% of the trustees) or, over the trustees’ objection, a stacked consent of continuing founders, Class E, and Class A. The Wall Street Journal’s reading of the objection path is: for three years after the IPO, unanimous continuing founders, a majority of employee shares voted, and a majority of Class A shares voted; thereafter, two-thirds of continuing founders, three-quarters of employee shares voted, and a majority of Class A voted; once Class F is gone, 85% of Class A outstanding. Those thresholds are charter and trust-instrument terms, not board policy.

What this means: employees are the swing constituency on one disputed Class T seat and on any attempt to kill the trust over the trustees’ objection. Public holders are not. After the founders’ special share sunsets, the trust still elects four of seven unless an extreme Class A supermajority ends it—and at that point the board itself becomes classified.

PBC status: statute, not a mission statement

Anthropic is a Delaware public benefit corporation. The amended and restated certificate of incorporation identifies the specific public benefit—the “Company Purpose”—as “the responsible development and maintenance of advanced AI for the long-term benefit of humanity.” Under DGCL §362, the corporation “shall be managed in a manner that balances” stockholders’ pecuniary interests, the best interests of those materially affected by its conduct, and that purpose. Directors are obligated, not merely permitted, to do that balancing. Under §365, a balancing decision meets the statutory duty if it is informed and disinterested and not such that no person of ordinary, sound judgment would approve. The S-1 says Anthropic cannot guarantee a conflict would be resolved in stockholders’ favor, and that judicial precedent on PBC balancing at this scale is thin.

DGCL §366 and the bylaws require a stockholder report at least every two years on performance against the public benefit. The board sets the qualitative standards; there is no third-party standard. DGCL §367 limits who can sue to enforce the balancing duty once the stock is exchange-listed: holders of the lesser of 2% of outstanding shares or $2 million in market value.

The prospectus treats several commercial choices as illustrations of that duty, not as permanent charter bans: a usage policy against fully autonomous weapons and mass surveillance; a gated Mythos preview via Project Glasswing rather than a broad release; a decision not to build image and video generation models; and planned public-benefit share issuances. Those policies can be revised. The duty to balance cannot, short of a charter amendment that itself has to clear the control structure. The company also warns that PBC status makes a hostile bid easier to reject even if it would maximize short-term yield, because new directors must weigh purpose and affected stakeholders, not only price.

Charter versus policy here: PBC status, the Company Purpose, the tripartite balancing duty, the biennial report, and the §367 standing threshold are binding (statute plus certificate and bylaws). Statements that the company “cannot assure” it will achieve the purpose, that directors “may” forgo revenue, and the current usage rules are disclosure and policy. “Financially disinterested” trustees is a structural constraint of the trust; the claim that they have the right expertise is a selection policy, not a legal standard a Class A holder can enforce.

Nasdaq listing and the controlled-company exemption

Anthropic has applied to list Class A on Nasdaq Global Select under ANT. It says it is a “controlled company” under Nasdaq rules because the LTBT, as sole holder of Class T, generally has the power to elect a majority of the authorized board—not because of Founder LLC’s 50.1%. That status permits it to opt out of:

• a majority-independent board;
• a fully independent nominating and corporate governance committee;
• a fully independent compensation committee.

The binding fact is the qualification. The choice of which exemptions to use is policy, and the prospectus is explicit about the current choice: it does not expect to rely on the exemption except for the nominating and corporate governance committee, which will not be entirely independent. The board currently has a majority of independent directors; audit and compensation committees currently consist entirely of independent directors. If it ceases to be controlled and remains listed, the nominating committee must become fully independent within Nasdaq’s transition period. Nothing in the charter locks in the current “we only use one exemption” stance. Reuters notes the practical use of that exemption: Daniela Amodei sits on the nominating committee.

What this means: controlled-company status is a consequence of Class T, so it lasts as long as the trust’s board power lasts. Investors who treat “majority independent board” as a permanent listing protection are relying on a stated intention, not a charter covenant.

Anti-takeover provisions: mostly charter, timed to the sunsets

The certificate and bylaws that become effective immediately before the offering are written as a takeover deterrent, and the risk factors say so in the company’s own words: the structure is “likely to have the effect of limiting the likelihood of an unsolicited merger proposal, unsolicited tender offer, or proxy contest for the removal of directors” and may deprive stockholders of a premium.

Provisions the filing lists as becoming effective in the charter and bylaws:

• The multi-class structure itself, so Founder LLC can determine mergers, sales, and other stockholder-approved transactions for the foreseeable future.
• Before the Class F sunset, non-Class T vacancies can be filled by the board on notice from Founder LLC, or by Class A and Class F voting together.
• Before the Class T sunset, board size can be set only with approval of 80% of authorized directors.
• Special meetings may be called only by the board (majority of authorized directors), the chair, the CEO, the president, and, before the Class F sunset, Founder LLC—not by ordinary stockholders.
• Written consent is tightly limited before both sunsets and generally barred after both sunsets (Class E actions must always be at a meeting).
• Advance-notice procedures for Class A nominations and proposals.
• No cumulative voting.
• Blank-check preferred: up to 30 billion shares, terms set by 80% of authorized directors and, before the Class F sunset, only with Founder LLC’s consent.
• After the Class F sunset, the board becomes classified into three staggered three-year classes, and 66⅔% of Class A voting power is required to amend certain charter and bylaw provisions.
• Exclusive forum: Delaware Court of Chancery (or another Delaware state or federal court) for fiduciary and DGCL claims; federal district courts for Securities Act claims; the District of Delaware for Exchange Act derivative claims. These are bylaw terms. Stockholders are not deemed to waive federal securities law.

On DGCL §203 (Delaware’s business-combination statute): the company has elected not to be governed by §203 until the later of the Class F sunset and the Class T sunset. The certificate provides that §203’s restrictions apply immediately after that later sunset. So the classic statutory anti-takeover rule is off while the special shares exist—because those shares already block a takeover—and switches on once they are gone, alongside the classified board.

Charter versus policy: every item in that list is a charter or bylaw term (or a statutory election embedded in the certificate), not a board guideline. The narrative that this is “likely” to deter bids is risk-factor disclosure of the effect of those terms. The decision to opt out of §203 only until the sunsets is itself a charter election, not a policy that management can quietly reverse without an amendment.

What outsiders have said — and what ISS and Glass Lewis have not

No public ISS or Glass Lewis report scoring this S-1, or a meeting recommendation on ANT, turned up as of November 4, 2026. That is the normal sequence: proxy advisors publish company-specific recommendations when there is a shareholder meeting, not when a preliminary prospectus drops. Their standing benchmark policies are still the relevant benchmark. ISS’s 2026 policy generally recommends voting against directors at companies with multi-class structures and unequal voting rights, with a limited exception for enhanced voting rights of limited duration. Glass Lewis is similarly skeptical of dual-class structures and of post-IPO governance terms it views as adverse to shareholders. Anthropic’s Class F has no time sunset of the sort those policies treat as curative; it sunsets only if the founder group shrinks below three or the LLC ends the share. Class T’s removal bar, once Class F is gone, is an 85% Class A vote. On those policies’ own terms, this structure would be expected to draw negative director recommendations at the first annual meeting. That is an inference from published benchmark policy, not a retrieved ISS or Glass Lewis note on Anthropic.

Named reactions that are on the record:

• Aswath Damodaran called the combination of the PBC charter and the voting design “so deviously self-serving that it would put Kim Jong Un to shame,” and said it “effectively emasculates public investors.” He distinguishes it from Alphabet and Meta, where he has owned dual-class stock, because those structures do not also give directors a legal duty to subordinate returns.
• Mark Levine, New York City comptroller (pension assets he oversees are reported above $325 billion; the office told Reuters the funds had about $7 million of private-equity exposure to Anthropic), had already asked Dario Amodei on October 14 to pause the IPO over safety, governance, and pace. After the S-1, he told Bloomberg the filing raised new concerns and described “a very worrisome trend of disempowerment to shareholders and all the balance of power tipping towards management,” comparing it unfavorably even to SpaceX. He also said the city funds would likely still get index exposure.
• Gil Luria (D.A. Davidson) called the seven-person controlling bloc an “unusual permutation” versus the one- or two-founder model at Google, and said the PBC—directors deciding what “benefit of humanity” means—is the part of the S-1 that will take longest to absorb. Mark Mahaney (Evercore ISI) wrote that governance limits shareholder influence because the board must weigh mission alongside returns, the trust elects a majority, and the co-founders hold 50.1%.
• Law professors quoted by The Verge: Elizabeth Pollman (Penn) called it a unique structure “creatively designed for this founder-led and mission-driven company.” Dorothy Lund (Columbia) said Founder LLC means the founders effectively choose the three non-trust directors. Jens Dammann (Texas) called trust control of four of seven seats highly unusual and said founders plus the trust have “tremendous power” to prioritize safety over commercial pressure. Ann Lipton (Colorado) agreed that ordinary shareholders cannot select the board or make business decisions, then added the constraint the charter does not remove: Anthropic still needs capital, so market reality may force commercial pace even if the ballot box cannot.
• Lawrence Hamermesh, a Delaware PBC specialist, told CNBC the duty “to balance” is, in practice, “Good luck,” and that a company that must keep raising capital will still have to attend to investors—while the PBC adds “another dimension of uncertainty.”
• Jan Messerschmidt (Cohen Milstein) told Law360 that founder control fixed for as long as members hold 2% of IPO-date holdings is something he had not seen; “protection of shareholders will come from disclosure, liability and fiduciary duty law, not the ballot box,” and the fiduciary law here is “largely uncharted.”

What this means for anyone underwriting or competing: the sophisticated critique is not that the disclosure is thin. It is that the binding terms (Class F at a fixed 50.1%, Class T board majority, PBC balancing with a weak enforcement standing rule, §203 off until both sunsets, classified board after the founders leave) make the ballot box irrelevant, and that the “we will only use one Nasdaq exemption” and “trustees are the right experts” statements are intentions sitting on top of those terms. The investor protection that remains is securities-law disclosure and an unsettled PBC fiduciary standard, plus the capital-markets constraint Lipton flags: a company that must keep funding compute cannot ignore the price of its stock even if it can ignore its stockholders’ votes.

---

Recent Findings Supplement (November 2026)

Anthropic, PBC publicly filed its Form S-1 registration statement with the SEC on October 26, 2026 (amended November 2, 2026), making the full prospectus—including governance details—publicly accessible via EDGAR.[1][2]

This filing (under CIK 0002130357, file number 333-299633) details plans to list Class A common stock on Nasdaq under ticker “ANT,” with Goldman Sachs and Morgan Stanley as lead underwriters. The initial public offering targets the week of November 16, 2026 (subject to change), potentially one of the largest ever. The document spans hundreds of pages with extensive risk factor disclosures (over 90 pages in the initial filing) and describes a multi-class capital structure and governance framework designed to balance commercial growth with the company’s AI safety mission.[3][4]

Capital Stock Classes and Voting Control

Anthropic’s post-IPO capital structure uses five classes of common stock, with control concentrated in two single shares rather than dispersed economic ownership. Class A shares (sold to the public and most existing holders) carry one vote per share and full economic rights (pro rata dividends and liquidation). Class T (one share held by the Long-Term Benefit Trust) primarily enables election/removal of a board majority. Class F (one share held by Founder LLC, owned by the seven co-founders) carries 50.1% of total voting power on most matters submitted to stockholders generally; new issuances do not dilute this percentage. Class N shares (primarily for strategic partners like hyperscalers) have no voting rights except where Delaware law requires a class vote and carry economic rights similar to Class A. Class E shares (for non-founder employees) have limited voting rights on specified matters.[5][6]

These provisions, including the creation and rights of the special classes, are embedded in the Certificate of Incorporation (charter exhibits in the S-1, such as EX-3.1), making them binding charter terms amendable only by the required stockholder vote (subject to the existing control structure). Operational details around the LTBT and Founder LLC (e.g., trustee selection processes or internal voting within the LLC) appear in supporting documents like bylaws or trust agreements, functioning more as policy or contractual arrangements layered on the charter framework.[7]

This structure ensures founders retain decisive influence on ordinary matters while the trust oversees mission alignment via board control, limiting public Class A shareholders’ practical ability to affect outcomes regardless of their aggregate ownership.[5]

Board Composition, LTBT Authority, and PBC Framework

The board has seven seats. The LTBT (an independent body with financially disinterested trustees, currently chaired by Neil “Buddy” Shah and including Ben Bernanke, Richard Fontaine, and Christopher Stubbs, with one vacancy) holds the Class T share and elects/removes four Class T directors (currently Reed Hastings, Vas Narasimhan, Christopher Liddell, and a vacancy). The founders (via Class F) elect the remaining three directors. Daniela Amodei chairs the board; Dario Amodei also serves.[8][6]

Anthropic remains a Delaware Public Benefit Corporation (PBC). Directors must balance stockholders’ pecuniary interests, the best interests of those materially affected by the company’s conduct, and the specific public benefit purpose: “the responsible development and maintenance of advanced AI for the long-term benefit of humanity.” The LTBT is explicitly designed to institutionalize this mission focus. Certain LTBT actions (e.g., removal of a specified Class T director without cause) require nominating committee concurrence or an employee vote tiebreaker mechanism. The LTBT itself can be altered or terminated only by an 85% supermajority vote in some descriptions.[9][10]

Founder control via Class F sunsets after 10 years post-IPO or when fewer than three founders remain. The staggered board (three classes serving three-year terms) activates once founder control ends, serving as an anti-takeover measure.[11]

These board election and PBC balancing requirements are primarily charter-level (via the share classes and PBC designation), while specific trustee qualifications, internal LTBT procedures, and the employee vote mechanism function as policy or bylaw-level implementations.

Nasdaq Listing, Controlled Company Status, and Anti-Takeover Features

Anthropic qualifies as a “controlled company” under Nasdaq rules because the LTBT (via Class T) has the power to elect a majority of the board. This exempts it from requirements for a majority-independent board and fully independent compensation and nominating/governance committees.[5]

Anti-takeover provisions include the dual-control structure (LTBT board majority + founder voting power), the potential for a classified/staggered board, and standard Delaware charter provisions (e.g., supermajority requirements for certain amendments once control structures lapse). The S-1 explicitly notes that the governance setup may make it more difficult to replace directors or management and could deprive stockholders of opportunities to sell at a premium.[6]

These exemptions and provisions are tied directly to the charter-defined share classes and board election mechanics.

Risk Factor Disclosures and Governance Risks

The prospectus devotes extensive space (roughly 80–94 pages) to risk factors, far beyond typical boilerplate. Key governance-related risks include potential misalignment between LTBT/founder interests and public stockholders, limited influence for Class A holders, challenges in balancing PBC duties with profitability, and uncertainties in Delaware PBC fiduciary duties (described as “largely uncharted”). AI-specific risks encompass model misalignment (e.g., self-preserving behavior resembling blackmail or resistance to shutdown), assistance to malicious actors (bioweapons, etc.), loss of human control, and geopolitical destabilization.[12][10]

These disclosures appear in the risk factors section of the prospectus (not purely charter terms) and serve as forward-looking warnings rather than binding obligations.

Analyst, Expert, and Proxy Advisor Views

As of November 4, 2026, no formal reports from ISS or Glass Lewis have surfaced in public sources (the filing is very recent, and proxy advisor commentary typically intensifies closer to or after listing). Governance experts and media analyses highlight the structure’s novelty: it has “no direct public company precedent” and layers mission-driven oversight (LTBT) with founder voting control in a way that significantly insulates the company from short-term investor pressure while raising questions about accountability to public shareholders.[13]

Commentators note the fixed 50.1% founder vote (non-dilutable and without a traditional sunset in some descriptions) and the trust’s board majority as mechanisms that prioritize long-term mission stewardship but could complicate standard governance oversight. Some view it positively as a “bold but sane” innovation for an AI company; others flag entrenchment risks and the untested nature of PBC duties in this context.[10][10]

Implications for competitors or entrants: Replicating this exact dual-control + PBC + independent trust model would require careful charter drafting and likely face scrutiny from investors and proxy advisors seeking stronger public shareholder rights. The controlled-company exemption provides flexibility on board independence but signals concentrated control that may affect valuation or institutional demand.[5]

All details above derive from the October 26/November 2, 2026 S-1 filings and contemporaneous reporting; earlier drafts or leaks predate the public filing but are superseded by the official documents.

OpenAI and Anthropic both use a Delaware public benefit corporation (PBC) so directors can weigh a charter mission against shareholder returns. They lock that mission in opposite ways. OpenAI puts legal control in a nonprofit parent whose powers were written into attorney-general conditions. Anthropic splits control between an independent purpose trust that elects a board majority and a founder vehicle that holds 50.1% of the general vote. Neither design is “mission language.” Both are control designs. The historical record says control designs hold; purpose language alone does not.

OpenAI: nonprofit control, written as a regulatory condition

On October 28, 2025, OpenAI completed a recapitalization that the California and Delaware attorneys general had spent roughly a year forcing into a narrower shape than the company first proposed. The nonprofit, renamed the OpenAI Foundation, kept the sole power to appoint and remove directors of the for-profit, now OpenAI Group PBC. At closing the Foundation held about 26% of the PBC, valued by the company at roughly $130 billion. Microsoft held roughly 27% (about $135 billion, down from about 32.5% on an as-converted basis). Employees and other investors held the rest, about 47%. Microsoft is the largest outside economic owner and has no board-appointment right. Control and capital point at different parties. [1] [2] [3]

The attorneys general did not “approve” the deal. Delaware issued a statement of no objection conditioned on commitments; California signed a memorandum of understanding and said it would not sue. Those conditions are the enforceable core, not the PBC label:

• The Foundation keeps sole power to appoint PBC directors and power to remove them.
• On safety and security, PBC directors must consider only the charitable mission and may not consider stockholders’ pecuniary interests.
• The board-level Safety and Security Committee stays at the Foundation, not the PBC, and can require mitigations up to halting a model release even when internal risk thresholds would allow release. Its chair (initially Zico Kolter) sits on the Foundation board, not the PBC board, with full observation rights into PBC meetings.
• The PBC’s mission tracks the nonprofit’s and the OpenAI Charter. The company must give the AGs advance notice of major governance changes, meet with them on a set cadence, and stay and expand in California.
• Within a year, a second Foundation director was to serve only on the Foundation and as a non-voting PBC observer. [4] [5] [6] [7]

Commercial terms moved the other way. Microsoft’s IP rights run through 2032, including AGI models once an independent panel verifies an AGI declaration, and OpenAI committed to an additional $250 billion of Azure purchases. Microsoft gave up a right of first refusal on compute and can pursue AGI on its own. The nonprofit’s equity is large, but Microsoft’s economic and contractual position is large enough to matter even without a board seat. [2] [8]

A year later the structure is being tested, not just described. In November 2026 the Delaware Department of Justice sent an investigative subpoena tied to safety commitments made at the recapitalization, after a Hugging Face-related incident. That is the mechanism working as designed: the enforcer is a state AG representing charitable beneficiaries, not a minority shareholder. Whether the commitments have teeth is now an open enforcement question, not a charter question. [9]

Anthropic: a trust elects the board; founders keep the stockholder vote

Anthropic is already a Delaware PBC whose charter purpose is the responsible development and maintenance of advanced AI for the long-term benefit of humanity. The PBC form gives directors legal cover to balance stockholders, people affected by the company’s conduct, and that purpose. It does not decide who chooses the directors. Anthropic added a second layer in 2023: a Long-Term Benefit Trust, a Delaware purpose trust whose trustees are not supposed to hold Anthropic securities, holding Class T stock with no economic rights and the power to elect a board majority. [10] [11]

The October 26, 2026 S-1 (Nasdaq listing sought under ANT; not yet a completed IPO as of early November 2026) freezes a five-class structure:

• Class T (one share, Long-Term Benefit Trust): elects the Class T directors, a majority of the authorized board (four of seven). No economic rights.
• Class F (one share, Founder LLC of the seven co-founders): 50.1% of total voting power on matters submitted to stockholders generally. No economic rights. Continuing founders each have a vote on how that share is cast.
• Class A (public and existing holders): one vote per share and pro rata economics. Together with Class F, elects the other directors.
• Class N (strategic partners, including convertible positions associated with Amazon and Google): generally non-voting, same economics as Class A, converts to Class A on a public sale or a transfer to an unaffiliated holder at or below 9.99%.
• Class E (employees, excluding co-founders): limited votes on specified governance matters, including parts of the trust-amendment path. [12] [13] [14]

Because the trust can elect a board majority, Anthropic expects to be a Nasdaq controlled company. Neither the trust’s director power nor the Class F vote has a calendar sunset. Trustees are paid by the company (the S-1 disclosed about $8.4 million paid to the trust in the nine months ended September 30, 2026, and cash trustee fees not tied to profits). Recent trustees have included Neil Shah (chair), Richard Fontaine, Ben Bernanke, and Christopher Stubbs, with a seat vacant at filing. [15] [16]

The override is real but expensive. The S-1’s “Trust Amendment Threshold” lets the trust or Class T be changed, or Class T redeemed for nothing, only if:

• trustees favor the change: 80% of authorized board seats and 80% of trustees in office; or
• trustees oppose it and Class F is still out: for the first three years after the IPO, unanimous continuing founders plus a majority of Class E votes cast plus a majority of Class A votes cast; at any time, two-thirds of continuing founders, 75% of Class E votes cast, and a majority of Class A votes cast; or
• no Class F remains and trustees oppose it: 85% of outstanding Class A. [15] [17]

Public Class A holders therefore cannot fire the trust-elected majority, cannot outvote the founders on ordinary matters, and cannot unwind the trust without the founders and a large employee vote. In October 2026 Anthropic also committed shares to an Anthropic Public Benefit Fund (an LLC, not a 501(c)(3)): 5% now, plus discretionary additions as value rises, with employee pledges bringing dedicated shares toward roughly 20%. That is an economic dedication, not a control right. [18]

How the two AI designs differ under pressure

	OpenAI (post-Oct 2025)	Anthropic (S-1 structure, Oct 2026)
Who picks the board	Nonprofit Foundation, exclusively	Trust elects a majority; founders plus Class A elect the rest
Who wins a stockholder vote	Not the public float in any operational sense; Foundation control is structural, not a super-vote share	Founder LLC’s single Class F share, 50.1%
Mission legal form	PBC, plus AG-mandated rule that safety decisions ignore stockholder money interests	PBC balancing duty only
Hard stop on a model	Foundation Safety and Security Committee can halt release	No equivalent external halt right disclosed; trust acts by electing and removing directors
Outside capital’s lever	Microsoft ~27% economics and a long IP/cloud contract, no board appointment	Amazon/Google largely in non-voting Class N
Who can sue or unwind it	AGs can enforce the non-objection conditions; ordinary PBC derivative suits face the Delaware 2% / $2 million threshold and a statutory safe harbor	Trust override only at the amendment threshold above; same thin PBC derivative path for balancing claims
What has already been tested	Original plan to drop nonprofit control was blocked by AGs; a 2026 Delaware subpoena is probing whether safety commitments are being kept	Not yet public; the trust’s majority was phased in before the filing, so the design has operated privately, not under a public float

The practical difference is the identity of the residual controller. At OpenAI, the party that can fire the board does not get richer by shipping a risky model, and a state official can go to court if the nonprofit stops behaving like one. At Anthropic, the party that can fire a board majority is a trust of disinterested trustees, but the party that can still carry every ordinary stockholder vote is the founding group. A safety fight that is a board fight favors the trust. A safety fight that is a charter amendment, a financing, or a sale favors the founders unless employees and the public also line up. Stacking both is stricter than either alone, and stricter than a plain dual-class PBC. It is not the same thing as OpenAI’s regulator-backed nonprofit parent.

Delaware PBC law, clarified in July 2026 in Drakes Landing, does not require directors to maximize sale price. They must balance stockholder money, affected stakeholders, and the charter benefit, and they get a safe harbor if the decision is informed, disinterested, and not one no person of ordinary judgment would approve. Only stockholders holding 2% (or, if listed, the lesser of 2% or $2 million of market value) can sue to enforce that balance. Beneficiaries of the mission cannot. A PBC charter is a shield for directors, not a sword for the public. [19] [20]

What other mission and control designs have actually done

Company	Control mechanism	Mission enforceability	Shareholder remedies	What happened under pressure
Meta	Zuckerberg’s 10-vote Class B: about 61% of votes on roughly 13% of the equity. 2026 Class C non-voting dividend lets the company raise capital and lets him sell a slice without giving up votes. No meaningful time sunset.	None. Ordinary Delaware corporation. “Long-term vision” is a founder preference, not a legal duty to anyone but stockholders.	Public holders can propose; they cannot win. Fiduciary suits are the real remedy and are hard. Controlled-company exemptions available.	Metaverse spending on the order of tens of billions could not be stopped by owners of most of the stock. A majority of independent holders has repeatedly backed one-share-one-vote; founder votes defeat it. Class C plan was approved by his written consent, not a Class A vote.
Alphabet	Page and Brin jointly above 50% of votes on roughly 11–12% of equity via 10-vote Class B. Class C is non-voting, created so acquisitions and equity comp would not dilute control.	None beyond ordinary fiduciary duties.	Same pattern: one-share-one-vote resolutions have failed for more than a decade because founders vote them down. 2012–14 Class C fight settled with issuance limits, not one-share-one-vote.	Founder control has held through product failures, regulatory crises, and institutional campaigns. Investors who want a vote already priced that in, or they own Class C and have none.
Snap	Public Class A has zero votes. Founders’ Class C has 10 votes and, at IPO, about 88–89% of voting power. Control lasts until both founders die or they sell down past a charter trigger.	None. The prospectus’s theory was that founders need years free of quarterly pressure.	No director election, no Rule 14a-8 proposals, and historically a path to skip a full proxy statement because the registered class does not vote. Remedy is sell, or sue for a fiduciary breach.	By 2026 the stock was down about 75% from the 2017 listing. Activist Irenic could publish a turnaround plan and could not threaten a vote. Founder control held; the “long-term vision” justification did not produce a durable premium.
Warby Parker	Delaware PBC and certified B Corp, plus 10-vote Class B for co-CEOs Blumenthal and Gilboa (a 2025-proxy analysis put them near 58% combined). Unused non-voting Class C is authorized. Class B converts to Class A on October 1, 2031, or earlier on transfer, departure, or 12 months after death or disability.	PBC balancing duty and biennial benefit report. B Lab recertification is private and revocable. Charter public benefits: access to vision care and community impact.	Ordinary stockholders cannot outvote Class B until sunset. Balancing-duty suits face the 2% / $2 million PBC threshold and the safe harbor.	Mission and founder control have coexisted because the founders still want both. The sunset is the unusual part: control is scheduled to end, which Meta and Alphabet refused. Untested against a founder who wants to drop the benefit.
Lemonade	One share, one vote. Classified board. PBC since before the 2020 IPO. Charter public benefit is insurance products where charitable giving is a core feature.	Deleting the benefit, or merging into a non-PBC or a PBC without the same benefit language, needs two-thirds of outstanding voting stock. Giveback is still a board decision subject to insurance capital constraints, not an automatic dividend.	Derivative suit to enforce balancing requires 2% ownership. Classified board and plurality elections slow a board fight.	The mission has not been formally unwound. It also has not stopped the board from sizing Giveback to regulatory capital. Soft, charter-level lock with real shareholder votes.
Coursera	One share, one vote. Classified board; directors removable only for cause by 66⅔%. Converted to a PBC in February 2021 before the IPO. Also a B Corp.	Same Delaware PBC machinery: balancing duty, benefit report, narrow derivative standing. No founder super-vote.	A majority of voting power elects the board over time. Supermajority only for removal for cause, not for ordinary strategy.	In the announced Udemy combination, the surviving company was set to keep the Coursera name and PBC status. Mission survived a change-of-control negotiation because both sides agreed to keep it, not because outsiders were legally barred from dropping it.
Patagonia (private)	2022: Purpose Trust took 100% of voting stock (about 2% of shares). Holdfast Collective, a 501(c)(4), took 98% non-voting stock and the profits not reinvested. Family guides the trust and still sits on the board. Cannot be taken public or sold in the ordinary way without the trust.	Voting control sits with an entity whose only job is the purpose. Economic residual claimant is an environmental nonprofit, so “shareholders” are not trying to extract the mission. B Corp status continues.	No public float. No activist can buy a block and demand a sale. The constraint is structural, not cultural.	Through early 2026 the Collective had received on the order of $100 million-plus in cumulative dividends (company had projected about $100 million a year, depending on results) and deployed capital into land and advocacy. Strongest lock in this set, because Chouinard gave the company away. Not available once outside investors already own the residual claim.
Danone	One share, one vote French listed company. In June 2020 shareholders approved société à mission status (PACTE law) by about 99%. A mission committee and an independent auditor monitor statutory objectives. No dual class.	The status survived the CEO. French law requires stated objectives and a mission committee; it does not let the mission committee fire management or block a board that wants a more conventional CEO.	Activists with a few percent, plus a board that lost confidence, were enough. Bluebell and Artisan did not need a super-vote.	Emmanuel Faber, the executive most identified with the mission, was removed as CEO and chair in March 2021 after underperformance versus peers. The société à mission status is still in the bylaws and still produces mission-committee reports (the 2024 report was the fifth). Purpose law did not protect the purpose-driven CEO. Financial recovery under Antoine de Saint-Affrique came with the mission language left in place and the governance fight already lost.

Sources for the rows above include Meta and Alphabet voting descriptions and the 2026 Class C plan, Snap’s zero-vote structure and later activist impasse, Warby Parker’s charter and conversion triggers, Lemonade’s charter thresholds, Coursera’s one-vote charter, Patagonia’s 2022 transfer, and Danone’s 2021 ouster plus later mission-committee reporting. [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34] [35]

What has held up

Purpose statutes protect directors. They do not protect the mission from the people who elect directors. Danone is the clean case. Société à mission was adopted almost unanimously and is still on the books. The CEO who embodied it was removed within nine months by ordinary shareholder pressure and a board that blinked. Delaware’s PBC statute does the same job more explicitly: it displaces Revlon-style price maximization and then makes enforcement expensive and shareholder-only. Lemonade and Coursera show the mild version. The charter can raise the vote needed to delete the benefit (Lemonade’s two-thirds is meaningful), and a classified board slows a raid. Neither stops a determined majority from replacing management, resizing a giveback, or negotiating the mission away in a merger if both sides agree. Coursera keeping PBC status in the Udemy deal is a choice, not a lock.

Dual-class locks a person, not a purpose. Meta and Alphabet are the proof that 10-vote stock survives product failure, political heat, and a decade of one-share-one-vote proposals. Snap shows the endpoint: if the public has no vote at all, even the ritual of an annual meeting stops being a constraint, and a 75% drawdown does not create a ballot. Warby Parker is the hybrid worth watching. It pairs a PBC and a B Corp with founder super-votes, but it also has a hard sunset on October 1, 2031. That is the feature Meta’s 2026 Class C plan was designed to avoid. A sunset converts “founder vision” from a permanent exemption into a term appointment. None of these structures failed to hold control. They failed, when they failed, at the thing dual-class is often sold as: a guarantee that the controlled strategy was the long-term one shareholders would have wanted.

The designs that bind are the ones that move residual control away from people who profit by abandoning the mission. Patagonia is the extreme and, so far, the most stable. Voting stock sits in a purpose trust; the economic stock sits in a nonprofit that exists to spend the dividend on the planet. There is no outside holder who gets richer if the company is sold to a buyer that drops the charter. The cost was giving the company away, including a structure built to avoid a giant tax bill that would have forced a sale. That option closes once venture investors own the residual claim.

OpenAI and Anthropic are attempts to get some of that separation without a Patagonia giveaway. OpenAI’s version is external: a nonprofit parent with appointment and removal rights, a safety committee that can stop a release, a statutory instruction to ignore stockholder money on safety decisions, and two attorneys general who have already shown they will withhold non-objection and who have subpoena power afterward. The weakness is commercial, not formal. Microsoft’s 27% stake and through-2032 IP rights mean the largest economic partner can shape what “safety-compatible commercialization” looks like without sitting in the appointment seat. The 2026 Delaware subpoena is the first public sign that the AG conditions are a living constraint rather than a closing memo.

Anthropic’s version is internal and layered. The trust is closer to Patagonia’s purpose trust than to a nonprofit parent: disinterested trustees, no economics, power to seat a board majority, and an override that requires founders, employees, and the public together, or 85% of Class A if the founders’ share is gone. The Founder LLC share is closer to Meta: control that does not dilute as the company issues stock. The combination is stricter than Warby Parker’s PBC-plus-dual-class, because a founder majority of the stockholder vote still does not elect a majority of the board. It is weaker than OpenAI’s AG package on one axis: there is no outside official whose job is to sue if the trust goes quiet, and no committee with an explicit halt right sitting outside the PBC. It is stronger on another: Microsoft-scale partners are largely in non-voting stock, so they cannot carry a vote even if they wanted to.

For anyone designing the next one, the pattern is consistent. Write the mission into a PBC or société à mission charter if you want directors protected when they trade short-term price for the stated benefit. Do not expect that sentence to survive a board the residual claimants still elect. If the goal is to survive a bad year, a founder fight, or a strategic investor with a cloud contract, put the power to fire directors in an entity that does not get paid for abandoning the mission, make unwinding that entity a supermajority that founders cannot deliver alone, and if the nonprofit route is used, expect a regulator to be the only party who will actually enforce it. Dual-class without a sunset has held under as much pressure as any structure in this set. It has held the founder. That is a different product from a mission.

---

Recent Findings Supplement (November 2026)

Anthropic's October 2026 S-1 filing introduced the most significant new details on its hybrid governance structure since its 2023 Long-Term Benefit Trust (LTBT) creation.[1][2]

The filing (Oct 26, 2026) for a Nasdaq listing under ticker ANT describes a Delaware Public Benefit Corporation (PBC) with a "novel capital and corporate governance structure" using five share classes to separate economic rights from control while balancing stockholder interests, stakeholder impacts, and the stated public benefit ("responsible development and maintenance of advanced AI for the long-term benefit of humanity").[3]

• Class structure mechanics: Class A (public investors) gets 1 vote/share and pro-rata economics; Class N (strategic partners like Amazon/Google) has economics but minimal/no voting rights except where required by law; Class T (one share held by LTBT) elects/removes a board majority (Class T directors, initially 4 of 7 authorized seats) with no general voting or economics; Class F (one share to Founder LLC, controlled by the 7 co-founders) carries 50.1% of total voting power on general matters (including non-T director elections and most proposals) with no economics—the vote is not diluted by new issuances; Class E (employees, excluding founders) has limited voting on specific governance matters.[4][5]
• LTBT trustees (e.g., Ben Bernanke, Richard Fontaine, Neil Shah, Christopher Stubbs; financially disinterested experts) are self-perpetuating and independent; one board seat has safeguards (nominating committee concurrence or Class E vote for certain changes).[6]
• Anthropic qualifies as a Nasdaq "controlled company" (exemptions for board independence and committees) due to LTBT board power. Founders retain control "for the foreseeable future"; sunsets apply later (e.g., Class F when few founders remain; potential staggered board post-Class F).[7]
• Additional elements: 5% of shares (plus escalators on valuation milestones) to a Public Benefit Fund; founders pledged 80% of personal equity to charity.[8]

This layers PBC legal duties, an independent mission-focused trust for board majority, and founder voting supermajority—explicitly designed to insulate from investor pressure while enabling IPO capital.[9]

OpenAI's 2025 recapitalization (nonprofit OpenAI Foundation controlling OpenAI Group PBC) saw stake dilutions and ongoing AG oversight in 2026, with no major structural shifts.[10]

• Post-Oct 2025 closing: Foundation held ~26% equity (plus warrant) and appoints/removes the PBC board; Microsoft ~27% (diluted to ~24-25% by March 2026 $122B round and other factors); employees/investors the rest.[11]
• April 2026 partnership amendments with Microsoft adjusted revenue sharing (capped), cloud flexibility, and IP terms through 2032 (non-exclusive in some areas).[12]
• AG conditions (primarily 2025 MOUs from California and Delaware, with monitoring continuing): Nonprofit retains sole board appointment/removal power over PBC; identical missions; PBC directors must prioritize mission (disregard pecuniary interests) on safety/security issues; Safety & Security Committee stays under nonprofit with authority to halt model releases; majority-independent PBC board; quarterly AG oversight; commitments to remain in California and use charitable assets appropriately.[13][14]
• November 2026 development: Delaware DOJ issued an investigative subpoena to OpenAI (first compulsory action post-recap) following a Hugging Face-related incident, probing safety commitments and mission compliance under its authority over the nonprofit beneficiary.[15]
• October 2026 state AG lawsuits (multiple states) targeted ChatGPT safety/child safety claims under consumer protection laws, seeking injunctions and remedies like data deletion—not directly altering governance.[16]

Recent updates on other mission-driven or dual-class companies are limited and do not show structural changes.[17]

• Danone (Société à mission since 2020): April 2026 Mission Committee report covered 2025 progress on health/environment/social objectives (aligned with UN SDGs and B Corp); 2026 AGM approved all resolutions including governance renewals and 2025 results; leadership restructure (from 5 to 3 geographic segments effective 2026) for agility, with mission intact.[18][19]
• Patagonia (steward-ownership via Purpose Trust for voting control + Holdfast Collective for economics since 2022): No new 2026 governance developments reported.
• Dual-class examples (Meta, Alphabet, Snap, etc.): No recent mission-specific updates or pressure-test outcomes in sources.
• Broader steward-ownership discussions (e.g., Bosch, Orval) emphasize decoupling control from economics via trusts/foundations, but lack new AI-specific or enforcement data.[20]

Comparative table (drawing from Anthropic S-1 and OpenAI recap details as primary new sources; other companies via established mechanisms referenced in context):

Control Mechanism

- Anthropic: PBC + LTBT (Class T, majority board election) + Founder LLC (Class F, 50.1% general vote) + limited public input.[1]

- OpenAI: Nonprofit Foundation appoints/removes all PBC directors (post-2025 recap).[11]

- Dual-class (e.g., Meta): Founder supervoting shares for board/control.

- Mission entities (Danone Société à mission; Patagonia steward-ownership): Legal purpose in bylaws + mission committee/trust oversight; Patagonia uses Purpose Trust (voting) + nonprofit (economics).

Enforceability of Mission

- Anthropic: Delaware PBC duties (balance stockholders, stakeholders, purpose) + LTBT independence + charter provisions; founders as "stewards."

- OpenAI: Identical missions + AG MOUs requiring mission priority on safety; nonprofit oversight of Safety Committee.[14]

- Dual-class: Primarily contractual/founder discretion (weaker legal mission tie).

- Danone/Patagonia: Statutory mission reporting/audits (Danone); trust legal structure (Patagonia).

Shareholder Remedies

- Anthropic: Limited by control (public Class A has minimal influence); PBC fiduciary duties allow suits but controlled board/trust structure constrains; Nasdaq controlled-company status reduces some protections.

- OpenAI: AG enforcement (monitoring, potential suits); nonprofit beneficiary rights.[15]

- Dual-class: Standard fiduciary suits, but founder control often prevails.

- Danone: Mission committee reports + auditor; limited direct stakeholder suits. Patagonia: Trust/nonprofit standing.

Outcomes/Pressure Resilience (to date)

- Anthropic: Pre-IPO; designed for longevity beyond founders (sunsets, staggered board potential); no public tests yet.[2]

- OpenAI: Implemented 2025 with AG concessions; 2026 probe/suits test safety commitments.[15]

- Dual-class: Proven for founder retention (Meta, etc.) but mission drift risks under investor pressure.

- Danone/Patagonia: Ongoing reporting/activism; Patagonia emphasizes permanence via fractionated trusteeship.[21]

Lessons from recent developments emphasize layered, independent mechanisms over single-point controls for resilience.[22]

Anthropic's combination (PBC legal duty + independent LTBT board majority + founder voting bloc) creates checks that neither pure dual-class nor standalone PBC achieves alone, reducing takeover or drift risks while attracting capital—though it limits public shareholder influence and relies on trustee/founder alignment. OpenAI shows AG MOUs can add enforceable guardrails (e.g., safety vetoes, oversight) post-restructuring, but invite ongoing regulatory scrutiny (as in the 2026 Delaware probe).[13]

Pure dual-class excels at founder control but offers weaker mission enforceability without additional duties or trustees. Société à mission or steward models (Danone, Patagonia) provide statutory or trust-based accountability via reporting/audits but face enforcement gaps (e.g., auditor selection biases or limited stakeholder standing) unless augmented (as Patagonia does via economic control transfer).[21]

For entrants or competitors: Replicating Anthropic-style hybrids requires careful charter drafting and independent trustees; AG involvement (as in OpenAI) can strengthen enforceability but adds compliance costs and litigation exposure. Designs holding up best appear to combine legal mission mandates, independent oversight bodies, and concentrated (but mission-aligned) voting—rather than relying solely on founders or standard corporate forms. No new empirical studies on long-term outcomes emerged in the period.