On October 18, Anthropic announced the Anthropic Public Benefit Fund, and Linas Beliunas called it “its own version of Universal Basic Income”. His post has 750 reactions and over 200 comments, and many of the comments tell him that is not what the announcement says.

I went and read the announcement itself. If you buy AI, my reading is short. It is a promise about Anthropic’s shares. Your contract is a different document, and the fund is not in it.

What the announcement says, and what it leaves out

The fund starts with 5% of Anthropic’s shares, “subject to stockholder approval”. The board can add 2% for every 50% increase in the company’s value above a baseline that has not been published. Counting what the co-founders and employees already pledged, Anthropic puts the total near 20% of its shares.

The fund is “a wholly-owned LLC”, chosen “rather than a 501(c)(3) or other format”, to allow “adjustability between different forms of giving”. None of the equity it holds will go to lobbying or political giving.

Now the gaps. Universal basic income does not appear in the text. On uses, the wording is that “we might explore” large donations to traditional nonprofits or direct distributions to specific populations. No amounts, no recipients, no dates. On governance, Anthropic says it will share more “in the coming months”.

One commenter argued the company could use the fund to pay for the safety of its own models. The announcement points the other way. Work that has “both a public benefit component and for-profit value”, like safety and interpretability research, is kept “directly in the company”. What the fund may do later is not addressed, and I do not know either.

I am not taking a side on Anthropic’s motives. A pledge with no published governance cannot be believed or dismissed yet. It can be waited on.

What you can verify today

For a buyer, the useful split is between what is written and what is intent.

Item	Where it lives	Status
Public Benefit Fund	Oct 18, 2026 announcement	Intent. Governance and milestones unpublished
Anthropic as a public benefit corporation	Long-Term Benefit Trust page	Published structure
Long-Term Benefit Trust	Same page	A trust holding stock that elects and removes part of the board
Commercial terms	Anthropic Commercial Terms	The text you sign

On the trust, Anthropic’s own page describes an independent body of five trustees holding Class T stock, with the power to elect and remove a share of the board that grows over time. That is a structure with powers you can read about. A fund with no rules yet is a different animal.

The public Commercial Terms, which I read in full for this post, say checkable things:

• “Anthropic may not train models on Customer Content”.
• An indemnity for third-party IP claims (section K).
• Either side can terminate for convenience, and Anthropic must give 30 days’ notice (I.2.a).
• Anthropic may assign the contract to an affiliate or in a sale of “all or substantially all its business”. You cannot do the same without consent (M.4).
• I found no service level agreement. On a suspension, Anthropic commits to “reasonable efforts” to give written notice (I.3.b).

Those are the standard published terms. Your negotiated agreement may say something else, which is a reason to open it.

What I would do with this

I use Claude Code every day, so I have a stake and I am saying so. I would run the same check on any vendor.

First, take the fund out of the decision matrix. There is nothing in it to score. Second, put a date on the calendar for when Anthropic publishes the structure and governance, and that week read who appoints whom and what gets reported to whom. Third, check the assignment clause in your own contract. If the company changes hands, that paragraph decides what still binds.

Fourth, ask about incident notice and availability, where the public text is thinnest. The six governance questions for your AI vendor work for any provider. And the Anthropic copyright settlement is a reminder of why an indemnity gets read slowly.

I have a few open questions I would like answered. Who decides how the fund is used? Will there be public reports? What is the baseline? When do stockholders vote? I don’t know, and neither does the announcement.

In the discovery phase of AI Maestro, each process ends in a Go or No-Go. If I am comparing vendors with you, a review of their terms comes before that call, on a two-column sheet: what is signed, and what is a promise. Write to us if you want that sheet built for the vendor you have today, Anthropic or not.

Build the terms sheet for your AI vendor