Anthropic's Long-Term Benefit Trust: What the WSJ Reports vs. What Is Public

The Wall Street Journal has published a look at Anthropic's Long-Term Benefit Trust, reported by Theo Francis and surfaced on Techmeme on October 19, 2026. Its framing, per the headline summary: an unusual, opaque five-member trust whose trustees own no equity and hold enormous sway over the company, with Anthropic sharing few details on how the trust's dynamics work. The article is paywalled and we could not read it, so this post does not describe its reporting beyond that summary. What we can do is lay out what Anthropic itself has published, what other outlets report, and what is still not public.

This is a "what is claimed and what is verified" post. If you want the short version: the broad design is public, the day-to-day behaviour of the trust is not.

TL;DR: the trust at a glance

What has Anthropic actually said about the trust?

Anthropic's own description calls the trust "an independent body of five financially disinterested members" with backgrounds in AI safety, national security, public policy and social enterprise. It is a purpose trust under Delaware law, and its purpose matches Anthropic's: balancing stockholder financial interests with the interests of people affected by the company's conduct, and its public benefit purpose. Anthropic is also a Delaware public benefit corporation, a distinction we unpack in our comparison of how AI labs structure public benefit.

The mechanism is a class of stock held only by the trust. Per Anthropic, it gives the trust the power to elect and remove a growing share of the board, phased in by time and funding milestones, with a stated plan for the trust to elect a majority of the board. A separate director seat is elected by Series C and later investors. The trust must also receive notice of certain actions that could significantly change the company or its business.

Trustees serve one-year terms, and new trustees are chosen by a vote of the existing trustees. Anthropic's July 2026 announcement adds that selection happens "in consultation with the company", which is the detail critics focus on: the body meant to check the company helps pick its own successors with input from the company.

Anthropic itself describes the structure as an experiment, and says it is not yet ready to present it as a model for others.

Who sits on the trust today?

Round table with empty chairs, showing how Anthropic's Long-Term Benefit Trust has shrunk from five trustees to three

Membership has moved a lot, and the footnotes on Anthropic's page track it:

• Jason Matheny left in December 2023 and Paul Christiano in April 2024.
• Richard Fontaine joined in May 2025.
• Kanika Bahl and Zach Robinson left in January 2026; Mariano-Florentino Cuéllar joined that month.
• Ben Bernanke, the former Federal Reserve chair, joined on July 9, 2026, bringing the trust to four alongside chair Neil Buddy Shah, Fontaine and Cuéllar.

In the announcement Bernanke said how AI's potential plays out "will depend, in part, on the institutions we build around it", and Shah said "the institutions built around this technology will matter as much as the technology itself", both per Anthropic's announcement.

A TechTimes report says Cuéllar stepped down on August 4 to become Anthropic's first Chief Global Affairs Officer, leaving three trustees and no announced successor. We did not find a second outlet confirming the exact date, so treat it as reported. The move itself is notable: a trustee who is supposed to be independent of management went into management, which is the sort of detail the WSJ's "opaque" framing points at.

What is the trust's "sway" in practice?

This is where the claimed-versus-verified line matters most.

Verified from Anthropic: the trust holds Class T stock, can elect directors, receives notice of major actions, and advises the board and leadership on major decisions, especially on AI's risks and societal impacts.

Reported, not confirmed by Anthropic in what we read: TechTimes says Anthropic has seven board seats, that the Class T shares carry no economic value, and that the original charter's milestone schedule leads ultimately to a trust-elected majority. A third-party wiki summary also claims the trust had appointed only one director as of late 2025, but we could not tie that to a primary document.

Not public as far as we found: the full trust agreement, how many seats the trust has actually appointed today, how often it has used its notice rights, whether it has ever objected to a board decision, and whether the milestones have been met.

That gap is the story. A body with "enormous sway" on paper is only as strong as its documented use, and none of the public material shows an instance of the trust blocking or changing a decision. Absence of evidence is not evidence of weakness, since a trust may shape decisions quietly through advice. But outsiders cannot tell the difference.

The failsafe: can stockholders override the trust?

Anthropic's page says stockholders can change the trust's structure and powers without trustee consent through "failsafe" provisions, but only with large supermajorities, and those thresholds rise as the trust's power phases in. Most changes are expected to come from agreement between trustees and the board or other stockholders.

Critics read this as a limit on independence: the entity that is supposed to constrain the company can, in the end, be amended by the people it constrains, if enough of them agree. Defenders read it as a necessary safety valve for a body whose purpose could otherwise become unworkable. Both readings fit the public text. What matters for investors is the actual threshold numbers, which would need to be disclosed in an offering document.

Why it matters now: the IPO

Anthropic reportedly filed a confidential S-1 on June 1, 2026 (see our S-1 explainer and the investor meetings coverage). Governance has become a live public question: New York City Comptroller Mark Levine asked Anthropic to pause the IPO until safety and governance concerns are addressed. Anthropic has also announced a Public Benefit Fund, and recent revenue figures are covered in our claimed vs. verified Q3 revenue post.

TechTimes also reports, citing The Information and Bloomberg, that Anthropic plans to give Dario Amodei and six co-founders a new class of stock with enhanced voting rights before listing, with details such as the vote ratio and any sunset undisclosed. Anthropic had not publicly commented in that coverage. If true, public shareholders would face two layers of control: founder votes and a trust-elected board majority. That is analysis from one outlet, and the author's own caveat is that the details stay unconfirmed until the prospectus is public.

What should the prospectus answer?

Glass box revealing gears, a picture of the disclosure investors need on Anthropic governance

If you follow AI governance, these are the concrete questions to look for when Anthropic's S-1 becomes public:

1. How many directors does the trust elect today, and what is the schedule to a majority?
2. What are the exact supermajority thresholds to amend the trust, and how do they change over time?
3. Who are the trustees now, how are they paid, and how are vacancies filled while the trust has fewer than five?
4. Does the trust have enforcement rights of its own, or only the stockholder-based ones?
5. Is there a sunset on Class T shares or founder super-votes?
6. Has the trust ever formally weighed in on a model release, safety policy or major transaction?

What this means for people who build on Claude

For most developers the trust changes nothing day to day. It does bear on a question builders increasingly ask: how much weight to put on a lab's safety promises when deciding where to build. Our coverage of Anthropic's independent evaluators and of the FTC probe into AI safety at OpenAI and Anthropic shows the same pattern: external checks exist, but their independence is the thing to examine. If you run agents in production, governance at the lab does not replace controls you own. AgentBeam is the agent security platform from the explainx.ai team that stops AI agents before they take dangerous actions.

How to check this yourself

• Read the trust announcement and its footnotes, which Anthropic updates when trustees change.
• Read the Bernanke announcement for the current wording on equity and selection.
• Watch for the public S-1 on the SEC's EDGAR site; governance sections and the charter exhibits are where thresholds appear.
• Read the WSJ article directly if you have access, and compare its claims with the list above. We have not.

Related reading

• How AI labs structure public benefit: LLC, foundation, PBC
• Anthropic S-1 and IPO explained
• NYC Comptroller asks Anthropic to pause its IPO
• Anthropic Public Benefit Fund: 5% of shares
• Anthropic hiring, pay and mission
• Anthropic's independent evaluators

Details reflect public sources as of October 19, 2026; trustee counts and IPO plans can change.