Late on 21 October 2026 the artificial-intelligence company Anthropic put a new document before the public. It is not a balance sheet, though balance sheets are coming. It is a kind of confession of intent, the first of what the company says will be an annual series, and it arrives at the moment when a private enterprise is preparing to sell pieces of itself to strangers.
The charter page Anthropic published opens with the company’s founding sentence: its charter is “to responsibly develop and maintain advanced AI for the long-term benefit of humanity.” Anthropic is a Public Benefit Corporation, a legal designation that obliges it to weigh what its shareholders want against what is good for everyone else. The new 2026 Public Benefit Report is the company’s own account of how it is doing that weighing, and of the work it has not yet done.
A grammar of arrival
The report rests on five pillars, and each is worded as though the journey were already over. In the first, the technical problems of AI safety have been solved, and powerful systems can be built and released without severe, irreversible harm. In the second, safe practice is universal among AI developers everywhere, held in place by industry norms and enforced by regulation. In the third, democracies lead in AI and use it to serve their publics and keep the peace. In the fourth, AI brings unprecedented economic growth, and the wealth lets every person pursue a good life. In the fifth, the benefits reach people everywhere and the opportunities are distributed.
The technical problems of AI safety are solved.
It is a sentence in the present tense. It describes a country that does not yet exist.
New republics often write their constitutions this way, as if declaring a thing could hurry it into being. That is a reflection, not an accusation, and the company does not hide the distance. The report, as Benzinga quoted it, warns that “Keeping AI systems safe could become more difficult as their capabilities outpace our ability to oversee them”. The pillars are destinations. The road to them, by the company’s own reckoning, may be getting longer.
The price of virtue
The report puts a price on these intentions in shares. In her opening letter, Daniela Amodei, Anthropic’s co-founder and president, writes that, subject to stockholder approval, the company is setting aside at least 5% of its shares to pay for projects that serve the charter. Each co-founder has also committed to pledging 80% of their equity to philanthropy, and employees have made equity-donation pledges that the company will match. Taken together, Anthropic hopes about 20% of its shares will end up serving the public good.
There is a mechanism for growth as well. Al Jazeera reported that the plan allows a further 2 percent to be added “at the board’s discretion” for every 50 percent increase in the company’s value. The richer the firm becomes, the more it may give away, if the board agrees. Benzinga reported a separate commitment of $200 million to an Economic Futures Research Fund, which is meant to study how AI could disrupt labour markets and to test policy responses.
Amodei’s argument, as the charter page summarises it, is that safe and accountable AI is the condition for any real benefit, not a specialist’s concern on the side. She ends on a promise: “some of humanity’s greatest moments await us on the other side.”
The calendar
The timing matters. On 19 October, two days before the report appeared, CNBC’s Kate Rooney reported that Anthropic would put a 5% equity stake into a new public benefit fund ahead of its planned initial public offering in November. In September the company’s prospectus warned that AI could pose an existential risk to humanity. A few days before the report, it opened its systems to an outside watchdog, with limits.
Read in that order, the documents tell a story of their own. The company warns of catastrophe, invites an inspector in on its own terms, and then, as the doors of the exchange begin to open, publishes its ideals, carved into five pillars.
Here a reflection is in order, and it is only that. The Public Benefit Corporation is an odd kind of sovereign. The law asks it to balance two masters, the shareholder and the public, but the shareholder holds shares and a vote, while the public holds only a report. The public’s ledger is kept by the company itself, which decides what counts as progress and when to publish it. It would be wrong to sneer at the pledge of a fifth of a company’s equity. It would also be naive to forget that the ruler who writes the chronicle of his reign does not always choose the harshest chronicler.
The ordinary person, the one the fifth pillar promises to reach wherever they live, has no seat in this arrangement. Workers whose jobs the $200 million fund will study are subjects of the research, not its authors. Citizens of the democracies the third pillar wants in the lead did not vote on the pillars. They are told they will benefit. They are asked to wait.
The 5% still needs the stockholders’ approval, and the listing is planned for November. The charter page ends with a modest line, the only one written in the future tense: “We plan to add to this record over time.”