This call is cross-partisan, pro-market, pro-competition, endorses no candidate or party, and applies to all frontier AI labs.
This statement is a work in progress. The guardrails below are proposals, and we want to hear from you: share suggestions when you sign, tick the box to collaborate or join deliberations, and take part in the discussions and deliberations ahead. Add your name
Our asks
Anthropic’s board of directors, Long-Term Benefit Trust and founders: Pause the IPO for 90 days, and use that time to adopt steward ownership and the guardrails below.
Institutional investors: Do not take part in a conventional offering of any frontier AI lab without these safeguards and an adequate stewardship model in place. Profiting in the short term from a technology that threatens our country, economy, self-government and humanity is at odds with your fiduciary duty to manage your beneficiaries’ funds for their long-term wellbeing. Steward models can also enable comparable, sometimes greater, financial returns over a longer period.
Legislators and oversight bodies: Support the pause, and require these standards of every frontier AI lab, scaled to its size and capability.
Anthropic and independent philanthropic funders: Fund a representative civic assembly of experts and members of the public, convened immediately, along with independently run consultations on how frontier AI is governed. They should produce workable recommendations within the 90 days, to shape the final guardrails and charter for Anthropic and guide other frontier labs. Final decisions should be ratified through a representative citizens’ assembly, with the ratification process set during those 90 days.
Who controls Anthropic today
We petition to these people to accept the request for a 90 day delay, and to engage in open and public discussions around the points included here.
| Group | Members |
|---|---|
| The seven co-foundersHold 50.1% of the vote together, through a single class of founder shares held by their Founder LLC, and elect three of the seven directors. |
|
| Long-Term Benefit TrustTrustees with no financial stake in the company. They elect four of the seven directors, and choose their own successors. |
|
| Board of DirectorsSeven seats, one of them currently vacant. |
|
Sources: the founders’ letter in Anthropic’s IPO filing, Anthropic’s company page and Long-Term Benefit Trust page, and CNBC.
Why current governance isn’t enough
Anthropic has done more than most AI companies to build structures for responsible governance. It is a public benefit corporation, and its Long-Term Benefit Trust elects a majority of its board. We recognize that work. But it is still not sufficient oversight for a company, and a sector, that could threaten our democracy, our country, our economy and humanity. Anthropic itself calls the Trust “an experiment.”
Control still rests with a small number of people. Under the company’s IPO filing, its seven co-founders hold 50.1% of the vote through a single share (CNBC, Morningstar), and a handful of trustees choose most of the board and pick their own successors. How control passes on over time is unclear. A group this small could be pressured by partners, funders, investors or others to act against the public interest, and there is no formal way for the public to hold it to account, even though its decisions will affect all of humanity.
We’re talking about technology that could become more powerful than all the world’s nuclear weapons combined. Recent events have shown how depending on one leader, or a few people, leaves an organization and the public exposed. The wellbeing of humanity should not rest on the good graces of a small number of people. That weight is too much to put on anyone’s shoulders, when they also carry their own reputations, status, pressures, salaries and future job prospects.
The most failsafe check is a randomly selected group of citizens, large enough to reflect the values and interests of the country, with no financial stake in the outcome and far harder to pressure or corrupt than an elected body or a small group of leaders. The structures below are suggestions, and still evolving. They can balance power among Anthropic’s board, the Trust, government oversight and regulation, and citizens’ assemblies or advisory boards. We welcome your suggestions, and we ask that this be opened to public dialogue before the IPO, not after.
Three possible guardrails for Anthropic and every frontier AI lab
These are proposals. The primary suggestion here is to slow down and immediately gather a representative civic assembly, bringing experts together with people from across the country, to deliberate on these questions and develop stronger governance and stewardship of these powerful technologies. Its recommendations should shape the final guardrails and the company charter. These below are possible ideas that could strengthen that stewardship to improve our collective future as these technologies continue to evolve.
Steward ownership, with protections that outlast the founders
Frontier AI is too consequential to be owned like an ordinary stock. Every frontier lab should adopt, and should be required to adopt, steward ownership, in which a mission-bound foundation or trust holds lasting voting control.
Steward-owned companies can still be strong investments, and can still be listed on the public markets. The difference is stronger guardrails against serious risks and harms, so they can deliver substantially more good to the public, to investors and to every stakeholder. Novo Nordisk, A.P. Moller-Maersk and Carlsberg all trade publicly while foundations keep control of the vote, and in Denmark, foundation-owned companies make up about 70% of the stock market’s value. Frontier AI labs have a much greater impact on the US, democracy, and humanity, and should have even greater safeguards in place.
This is essential to fiduciary responsibility and long-term profitability, for Anthropic, for every AI lab, and for the broader markets and economy. Managed responsibly, these technologies could bring tremendous benefits, financial and otherwise, to our whole economy, our country and humanity. Deployed irresponsibly, they could lead to the collapse of our economy, our country and humanity, which goes against every tenet of fiduciary responsibility.
- Lasting control: a mission-bound foundation or trust holds a majority of the votes, while outside investors can still own shares and share in the profits.
- Safeguards that last: protections for elections, free expression and the public that don’t depend on today’s founders or today’s trust.
- Locked into the charter: these protections can’t be changed without the Oversight Council, the Citizens’ Assembly and a supermajority.
We ask: adopt steward ownership and these lasting protections before any listing.
Strengthen the LTBT into a more empowered independent oversight council
A cross-partisan council of experts, stronger than the current composition of Anthropic's Long-Term Benefit Trust. Its members should include former election officials and democracy professionals, free-speech and national-security experts, faith, community and business leaders, Indigenous representatives, technical auditors, and a guardian who represents the rights of nature. It should have full independence, binding access to information, authority to commission audits, and the right to report publicly.
Within the current 5-member LTBT structure, we encourage the 5th seat, currently vacant, to be filled by someone with the role, authority, and funding to conduct civic assemblies online and in person across the US on critical AI issues, and represent those findings and the broader public within the LTBT.
We ask: write the Council’s independence, powers and funding into the company charter.
A permanent Citizens’ Assembly, with a Civic Fund, and a seat on the LTBT and the Board of Directors
We place our greatest faith in upholding the collective values of our country, our Constitution, and our shared future in a representative group of American citizens. Chosen by lottery like a jury, and supported by cross-partisan and nonpartisan experts, they can deliberate on questions that shape the future of our country and of humanity. Research on citizen deliberation shows that ordinary Americans, given balanced information and time to talk, find common ground more readily than today’s political system allows.
Within the current 7-member Board structure, we encourage the 7th seat, currently vacant, to be filled by someone with the role, authority, and funding to conduct civic assemblies online and in person across the US on critical AI issues, and represent those findings and the broader public within the Board of Directors.
- Oversight of the Council: the Assembly ratifies Council members, reviews its work, and advises it on priorities.
- Control of the Civic Fund: each lab contributes a minimum of 5% of its revenues each year, plus twice its total political spending from the prior year. The Assembly directs two-thirds of the fund and the Council one-third, for nonpartisan civic purposes only.
- Built to last: members are paid and rotate regularly, supported by a permanent, independently funded professional staff.
We ask: fund the Assembly, its staff and the Civic Fund for the long term, and give the Assembly these roles in the charter, along with a seat on the LTBT and the Board of Directors.
Why this matters
These are hard questions, and they deserve careful debate. But if citizens have no real oversight of the most powerful tools ever built, we are on a path toward no longer governing ourselves, in the United States or anywhere else. We urge immediate consultations and citizen deliberations on these questions, so that we build a human-centered future that strengthens our democratic republic. This is about nothing less than the future of our country, our democracy, our economy and our humanity. It deserves careful attention and much stronger safeguards, so that we navigate this transition well. Nothing less is acceptable, for Anthropic or for any frontier AI lab.
Details and FAQ
Open any question for the details.
Why a 90-day pause, and why now?
Anthropic is expected to price its shares on November 18 and begin trading on November 19, 2026 (Yahoo Finance). Ownership and governance are far easier to change before shares are sold than after. Ninety days is short against the life of the company and long enough to put lasting safeguards in place.
Doesn’t Anthropic already have strong governance?
Anthropic has done more than most to build responsible governance, and it has taken real steps. It is a public benefit corporation, and its Long-Term Benefit Trust elects a majority of the board. But the Trust is small, chooses its own members without public input and reports little publicly, and Anthropic itself calls it “an experiment.” Under the IPO filing, the founders hold 50.1% of the vote, a fixed controlling stake that will one day sunset, with no clear plan for what follows (CNBC, BIT). Citizens have no formal role.
What about the founders’ pledge to give away 80% of their shares?
We appreciate it, along with the company’s own contributions (summary of the filing). But with stakes this high, how those funds are used should not depend only on the interests of a small number of people. Public engagement is critical to make sure they serve the broader, long-term public good. That could come through a mix of philanthropy, government, including substantially higher taxes on AI-generated revenue, through citizens’ assemblies taking part in the decisions, or a combination of all of those, or potentially other approaches.
Does this apply only to Anthropic?
No. These standards should apply to every company developing frontier AI, from large established technology companies to new labs, public or private. Anthropic is first because its offering is imminent.
Are you against markets or AI companies?
No. We support free markets, competition and American innovation. Our concern is concentrated, unaccountable power, whether in a few companies, a few investors or government. Obligations should scale with a company’s capability, revenue and reach, so they never become a moat that protects incumbents or shuts out new competitors.
What is steward ownership?
In steward ownership, a mission-bound foundation or trust holds lasting voting control, so the company can’t be sold or steered away from its purpose. Outside investors can still own shares and share in the profits.
Some of the strongest examples are publicly listed. Novo Holdings, owned by the Novo Nordisk Foundation, holds about 28% of Novo Nordisk’s shares but about 77% of the votes (Novo Nordisk). Foundations also hold majority voting control of A.P. Moller-Maersk and Carlsberg. Bosch, Carl Zeiss, Patagonia and the John Lewis Partnership are steward-owned and privately held.
Hershey, which is listed and controlled by a trust, is a reminder that the design details matter: its trust has faced governance controversies over the decades. Frontier AI built inside larger companies should be under equivalent governance.
Can a steward-owned company still be a good investment?
Yes. Foundation ownership is common on public markets. Researchers at Copenhagen Business School have mapped listed foundation-owned companies around the world, from Maersk and Hershey to Lindt and Tata Consultancy Services (CBS). In Denmark, foundation-owned companies make up about 70% of the stock market’s value (CBS). A 2024 study of 411 listed companies found that lenders give foundation-owned companies more favorable loan terms than companies controlled by families or institutional investors (summary).
With frontier AI, the stakes go well beyond any one company. Responsible deployment could bring tremendous returns across the whole economy, and irresponsible deployment could take down the markets those returns depend on. Steward ownership helps protect both.
Why should institutional investors hold back?
Pension funds, endowments, foundations, donor-advised fund sponsors, family offices and other long-term investors owe duties to their beneficiaries over decades. In our view, investing in a company and sector with risks this large, without adequate protections and lasting stewardship, runs against the prudence and long-term care that fiduciary duty requires. Charitable capital also has public purposes to uphold.
Which protections should outlast the founders?
At a minimum: a mission lock on safety, election and free-expression commitments; a strengthened, transparent trust with citizen input; independent audits of viewpoint neutrality; election-integrity safeguards against foreign interference and deceptive AI content; no surveillance of lawful political, religious or civic activity; open markets; full disclosure of political spending, lobbying and government contracts; whistleblower protection; a voice for the rights of nature; and no sale or change of control without Council and Assembly review.
Why trust a citizens’ assembly?
In the 2019 “America in One Room” project, a stratified random sample of more than 500 registered voters deliberated for a weekend with balanced briefings. Of the 26 most polarizing proposals, Democrats and Republicans moved closer together on 22 (APSR study via Stanford, summary). Randomly selected citizens’ assemblies have been used in many countries to work through divisive questions.
Why random selection, and not elected or appointed leaders?
A randomly selected group, large enough to reflect the country, is the most failsafe check we know of. Its members have no financial stake in the outcome and no campaign or career riding on it, so they are far harder to pressure, lobby or corrupt than an elected body or a small group of leaders. There is substantial research and experience with citizen assemblies, and they have demonstrated effectiveness in coming to cross-partisan solutions that politicians were not able to. Juries work this way for the same reason.
How would the Assembly be chosen and supported?
The pool combines self-nominations with random invitations by mail, so it isn’t limited to the already engaged. Members are selected to reflect the country by region, age, party, income and education, are paid, and rotate on fixed terms. A permanent professional staff provides neutral facilitation, expert briefings and administration.
How large is the Civic Fund, and what can it pay for?
Each year, each lab contributes a minimum of 5% of its revenues, plus twice its prior-year political spending, to offset partisan spending with nonpartisan civic strengthening. The 5% applies above a certain revenue size and is tiered, so larger companies may contribute more; the details are for future deliberations.
The fund supports only nonpartisan civic work and innovation, so that civic governance can advance at the pace of AI and other frontier technologies. That is the only way to ensure these tools have a positive future for humanity. It can fund civics education, voter information, election administration, local journalism and other areas, guided by (but not limited to) the Principles of our Democratic Republic. It never funds candidates or parties, and the company has no say over grants. Additional contributions may support economic transitions, retraining and other critical areas as we navigate the transition to an AI-powered economy.
Who speaks for the rest of the world?
Decisions made by frontier AI labs will affect all of humanity, yet no one outside a small group has a formal say. At a minimum there should be strong accountability to American citizens. We also think there should be a global governance component that represents the voices of people around the world.
Why include the rights of nature?
Keeping our home planet livable is core to everything else, and today’s governance gives the natural world no voice. There is already precedent: in 2022, the UK company Faith in Nature changed its constitution to give Nature a voting seat on its board, held by a guardian with a duty to represent it (Ecojurisprudence Monitor). Frontier AI labs could put the same kind of role in place, on the Oversight Council and the board.
Haven’t AI company leaders warned about these risks themselves?
Yes. In his January 2026 essay The Adolescence of Technology, Anthropic CEO Dario Amodei warned of civilization-level danger without timely intervention. He wrote that AI-enabled authoritarianism terrifies him and identified AI companies themselves as a major source of risk (Axios). We take those warnings seriously, and ask that they be matched with lasting, citizen-backed safeguards.
Is this partisan?
No. The statement is cross-partisan and endorses no candidate or party. We welcome signers from every political tradition who believe Americans should have a real say in how these technologies are governed.
Is this statement final?
No. This is a work in progress, and the guardrails here are proposals. A representative civic assembly of experts and members of the public should be convened right away to deliberate on them, and its recommendations should shape the final guardrails and charter.
We welcome your help. Share suggestions in the form when you sign, tick the box to collaborate or join deliberations, or email us at ai@democracyweek.net.