On Sunday, Anthropic said it is setting aside part of its stock for a new Anthropic Public Benefit Fund. If stockholders approve, 5% of the company's shares go in right away. After that, the board can add 2% more, at its discretion, for every 50% rise in the company's value above a baseline it has not published yet. The fund is a limited liability company (LLC) that Anthropic owns outright, not a charity, so it can move between gifts to nonprofits and direct payments to particular groups of people.

Anthropic's first example is workers whose jobs artificial intelligence (AI) may upend, on top of the $200 million it has already pledged to study that problem. None of the fund's shares can be used for lobbying or political giving, in the US or anywhere else. Counting its co-founders' pledges and an employee match program, the company says about 20% of its shares are now set aside for the public good. Details on governance, the baseline and the milestones are promised "in the coming months," and CNBC notes that the company's planned stock market listing is reportedly set for November. (Anthropic announcement; CNBC)

The obvious comparison is OpenAI. When it finished restructuring on October 28, 2025, its nonprofit became the OpenAI Foundation and took 26% of the for-profit OpenAI Group, worth about $130 billion at the time, plus a warrant for more shares if the company hits a valuation milestone. Microsoft holds about 27%. (OpenAI, October 2025 announcement)

The difference is control. OpenAI's foundation owns a slice of the company and also controls it. Anthropic's fund will hold shares, but the board is shaped somewhere else, by a separate trust. Both plans hand out more equity as the valuation climbs. Not everyone is sold: on X, Elon Musk answered Anthropic by suggesting it give the shares directly to the public. (CNBC)

Anthropic is a public benefit corporation, a Delaware form that changes what directors owe. Under the state's corporate law, such a company has to be run in a way that balances its stockholders' financial interests, the interests of people materially affected by what it does, and the public benefit written into its charter. At least every two years it must give stockholders a statement of the board's objectives, the standards it uses to measure progress, and facts on how it did.

Anthropic adds a layer on top. Its Long-Term Benefit Trust, five members with no financial stake in the company, can elect and remove a growing share of the board, eventually a majority, according to its 2023 description. The new fund comes with a promised public benefit report, and that report is where investors will look for the baseline and the milestones. A pledge that grows at the board's discretion is only as firm as the minutes that record it. (Delaware Code, benefit corporations; Anthropic on the trust)

Governments want a stake too, though theirs looks different. On Monday, as the Invest in Germany Summit opened in Berlin, the federal government announced plans for an AI partnership with Nvidia, and Chancellor Friedrich Merz was due to meet Nvidia chief executive Jensen Huang. About a dozen German companies are to take part, among them Siemens, Audi and the image start-up Black Forest Labs, to push AI into robotics, factory work and self-driving cars.

Euronews reports the federal government will put in a sum in the hundreds of millions of euros, mostly for computing time on Deutsche Telekom's industrial AI cloud, which runs on Nvidia chips. That is a public stake of another kind: money for capacity and a seat at the table, not shares. Huang, for his part, told reporters in London that Germany has fallen well behind in AI. (deutschland.de, in German; Euronews, in German)

Already filed, if you want the run-up: The Label (Indonesia drafting rules to label AI-made content), The Oath (New York's Council putting SpaceXAI under oath), and The Accord (the White House pact on frontier AI duties, independent auditors included).

Also on the shelf: The Brake (Satya Nadella on treating models as insider risks), The Probe (the Federal Trade Commission looking into OpenAI and Anthropic) and The Near Miss (London wanting AI near misses reported). Read today beside The Accord. That one was about who checks the labs. This one is about who shares in what they make.

Giving the company away is not new. In September 2022, Yvon Chouinard, Patagonia's founder, and his family transferred their ownership. All the voting stock, about 2% of the company, went to a new Patagonia Purpose Trust, and all the nonvoting stock, about 98%, went to a nonprofit called the Holdfast Collective. Profits not put back into the business would go to fighting climate change, about $100 million a year by the company's estimate at the time. (Patagonia, 2022 letter from Yvon Chouinard)

The split is the lesson. One holder got the votes and the job of guarding the mission, and the other got the money. Anthropic now has the same two pieces, a trust that shapes the board and a fund that holds shares. The twist is one Patagonia never needed: Anthropic's gift grows only if the company's value does, and the board decides when.

Put these to management. As AI changes our workforce, have we promised the people affected anything, in money, retraining or time, and who signed off on it? If we ever tied a commitment to our share price or valuation, who would set the baseline, who would decide when the next tranche goes in, and where would the board's vote be recorded?

If we buy from, or invest in, an AI company with an unusual structure, do we know who controls it, whether a foundation, a trust or ordinary stockholders, and what happens to our contract if control shifts? And if we are a benefit company, or thinking about becoming one, when did the board last read its own benefit statement? Generosity is easy to announce. The terms are what a board has to own.

The Move

AI companies are starting to write the public into their share registers, and they are doing it in different ways. OpenAI gave its foundation control and about a quarter of the company. Anthropic is giving a fund 5% now and more if it grows, with the board holding the dial.

Ask the general counsel and the chief financial officer to bring one page to the next board meeting: every promise we have made, or are thinking of making, about sharing AI's gains with workers, customers or the public, the trigger for each one, who decides, and how we would report on it.

If a promise has no trigger and no owner, it is a press release. Start there.

Seven bullets. One move.