Anthropic's 5% Public Benefit Fund explained: who gets the shares, and when

Anthropic, the maker of Claude, wants to give away a slice of itself: 5% of its shares now, and possibly more each time the company's value climbs. Priced at its last private valuation, the first 5% would be worth about $48 billion (our arithmetic: 5% of the $965 billion May valuation). Anthropic says it could be used on problems such as workers displaced by AI. The pledge arrives weeks before what could be the largest IPO in history, so how it works, and what it leaves out, matters well beyond Anthropic.

The plain-English version

On Sunday, 18 October 2026, Anthropic announced the Anthropic Public Benefit Fund. The idea is simple to state. The company sets aside 5% of its ownership for "the public good", and if the company keeps growing in value, it may set aside more.

Think of a family that owns an orchard and promises one row of trees to the village. They also promise to plant another row for the village each time the orchard grows by half. Nobody gets fruit today. The village gets a claim on future harvests, and the claim is worth more if the orchard thrives. Anthropic's fund works the same way. It doesn't receive cash. It receives shares, slices of company ownership, which can later be sold or used to pay for things.

Some details matter more than the headline:

• It's not final. Anthropic's existing stockholders still have to vote to approve it, and no date has been published for that vote.
• It's not a charity. Big corporate giving often goes through a registered charity, such as a company foundation. This fund is a company that Anthropic itself fully owns, a "wholly-owned LLC" in Anthropic's words. Anthropic says this gives it flexibility. It could make big donations to existing nonprofits, or give money directly to particular groups of people. The example Anthropic raises is the effect of AI on jobs.
• Some uses are ruled out. Anthropic says none of the fund's shares will be used for lobbying or political donations, in the US or abroad.
• Much of it is still blank. Who runs the fund, how growth will be measured, and when it starts spending are all promised "in the coming months".

Why now? Anthropic is preparing to go public, which means selling shares on the stock market. Our reading is that a large carve-out is simpler to approve now, while the company has a small group of private stockholders, than after thousands of public shareholders arrive. The company also frames the fund as a choice about pay. Anthropic says it is directing equity to the public benefit instead of the escalating executive compensation awards common at tech companies.

If you remember one thing: the fund is a promise of ownership, not a pile of money. It will be worth exactly what Anthropic's shares turn out to be worth.

How it actually works

The problem it's trying to solve

Anthropic's leaders have argued for years that their technology could reshape work. The company already put $200 million into an Economic Futures Research Fund to study the labor-market effects of AI. Its July agenda gives typical grants of $5–30 million and topics that include modernising income support and "building worker stakes in AI-driven growth". In its new announcement, Anthropic says that "given the pace of AI progress, bolder approaches may be needed" (Anthropic). A research budget can measure disruption. It can't do much to cushion it.

The old way

Tech companies usually give in one of three ways:

Anthropic already does the second and third. Its co-founders pledged 80% of their equity at founding. An employee donation-match programme has about half of staff participating. Discounted Claude access for nonprofits and research institutions is part of normal company work.

The new idea, step by step

1. Carve out ownership, not cash. Subject to a stockholder vote, 5% of Anthropic's shares go to the fund.
2. Add a growth escalator. For every 50% increase in Anthropic's value above a set baseline, the board may add another 2% (Anthropic). The analysis outlet Finimize calls this a "dilution kicker". The fund's share of the company rises with the company's success.
3. Keep it in-house. The fund is an LLC owned entirely by Anthropic. It is not an independent charity.
4. Fence off politics. The fund's equity cannot go to lobbying or political giving.
5. Spend flexibly later. Possible uses include large donations to traditional nonprofits and "direct distributions to specific populations" (Anthropic).

Why tie it to the share price?

The escalator links the size of the public's claim to the size of the bet the public is being asked to absorb. If Anthropic's systems turn out to be only modestly valuable, the company stays smaller, the disruption is presumably smaller, and the fund stays near 5%. If AI turns out to be enormously valuable, which is the same scenario in which job disruption would be largest, the company's value rises and the board can grow the fund. In finance terms, the public gets something like a claim on the upside of AI.

Anthropic says that counting the new fund, the founder pledges and the employee match, roughly 20% of its shares will be dedicated to the public good, plus any escalator top-ups.

What it costs, and who pays

If the fund's shares are newly created, every existing shareholder owns a slightly smaller slice. Anthropic's announcement doesn't say whether the shares will be new, where they come from, or how the fund will turn stock into spendable money. That is why the stockholder vote matters. Existing investors are being asked to give up part of their stake.

The deep dive

Where the fund sits in Anthropic's governance stack

The fund adds a fourth layer to an already unusual structure.

The first three layers decide how the company behaves. The new one decides where some of its value goes. The two kinds of control are separate, and the fund's own governance is the piece Anthropic hasn't published.

Two caveats apply to the table. The prospectus details come from Reuters' reading of a draft that Anthropic confidentially submitted and that is not yet public. They could change before listing. Also, Anthropic's LTBT page lists the Trust's original members and notes later changes: according to the page, current trustees include chair Neil Buddy Shah, Ben Bernanke, Richard Fontaine and Mariano-Florentino Cuéllar (Anthropic).

The escalator, formally

Let $V$ be the company's value and $B$ the undisclosed baseline. The fund's maximum target fraction is

where $k(V)$ is the number of 50% steps the board chooses to recognise. Anthropic's wording, "for every 50% increase in company value above a set baseline", allows at least two readings:

• Additive steps (each step is $0.5B$): $k = \lfloor (V - B) / 0.5B \rfloor$
• Compounding steps (each step is ×1.5): $k = \lfloor \log(V/B) / \log 1.5 \rfloor$

They give quite different funds:

Our arithmetic. Anthropic has not said which definition applies, or whether "2%" means 2% of shares outstanding at the time of each top-up.

In both readings the top-ups are discretionary. The board "may" add them. That makes the escalator closer to an option held by the board than to a contractual obligation, and it is why the yet-to-be-published milestones matter so much.

Dilution maths

Suppose the company issues new shares so that the fund ends up holding a fraction $p$ of the post-issuance total $N'$. Starting from $N$ shares:

An existing holder with fraction $s$ ends with $s(1-p)$. For $p = 5\%$ that is a 5% relative haircut, so 1.000% becomes 0.950%. A later top-up to a cumulative $p_2$ compounds the same way. If the shares instead come from an existing reserve, such as a pool that would otherwise fund executive awards (the framing Anthropic uses), the incremental dilution relative to the company's existing plans could be smaller. Anthropic's announcement does not say which applies.

What the 5% is worth: a range, not a number

Values are 5% × valuation, our arithmetic. Reported IPO valuations differ across outlets, some lower than $2T, and none is final.

For scale, the existing $200M Economic Futures fund is about 0.4% of the $48B figure.

Why an LLC, technically

A 501(c)(3) gets tax advantages, but it must follow charity law on how money is spent. "Direct distributions to specific populations", meaning cash to defined groups rather than grants to organisations, is the kind of activity that is awkward for a private foundation. A wholly owned LLC can do it, but gives up the tax advantages. Tech Insider notes that this structure is the reverse of OpenAI's: there, a nonprofit board has ultimate authority over the for-profit company, while Anthropic's giving vehicle sits below its corporate board.

(One legal caveat for experts: a wholly owned LLC isn't a separate taxpayer by default, and Anthropic hasn't said how the fund will be taxed or whether it might later route money through charities. "No charity tax treatment" describes the LLC itself, not every gift it might eventually make.)

There is a subtle consequence. While the LLC holds Anthropic stock, the "public benefit" money is economically still inside Anthropic's corporate family. Value only leaves when the LLC sells shares or distributes their proceeds. So the timing and governance of those sales decide when the public actually benefits, and Anthropic hasn't specified either.

Comparison with the closest prior commitments

• Founder pledges (80% of co-founder equity): these are personal and leave with the individuals. The fund is corporate and stays with the company.
• Economic Futures Research Fund ($200M): cash, focused on research, with a $1M minimum grant (Anthropic). The new fund is equity-based and can also make direct transfers.
• The fund's own agenda item: "building worker stakes in AI-driven growth", one of the Economic Futures priorities, could be read as a description of what an equity-funded vehicle is suited to do. That connection is our inference. Anthropic hasn't stated it.

The IPO backdrop

Anthropic announced on 1 June that it had confidentially submitted a draft S-1, and a listing is widely expected before the end of 2026. On 14 October, New York City Comptroller Mark Levine, who advises pension funds with more than $325 billion in assets, wrote to Dario Amodei asking Anthropic to pause the IPO until it addresses safety, governance and development-pace concerns. The letter also asked for specifics on audits, controls and a timeline. The fund doesn't answer those questions, which are about how the company builds AI, not where its value goes. It does show that the company is setting governance terms before public investors arrive.

Why it matters

For everyday people. Anthropic has explicitly tied a large, potentially growing claim on its own value to the possibility that its products disrupt jobs. (It isn't the only lab whose equity backs a public-interest body: OpenAI's nonprofit foundation also holds a large stake in its for-profit company. What's distinctive here is the valuation-linked escalator and the explicit labour framing.) If AI does push people out of work, an equity-funded pot that grows because AI is valuable is a different kind of cushion from a fixed research budget. Nothing reaches anyone yet, though, and Anthropic has not named recipients or a timeline (Tech Insider).

For developers and builders. Probably nothing changes directly. Indirectly, Anthropic already offers discounts on Claude for nonprofits and scientific institutions as normal company work. A large external fund could finance organisations that then buy AI tools and services, including from Anthropic's competitors.

For companies and investors. This is a cap-table event. Prospective IPO buyers will want to know the source of the shares, the baseline and the payout policy, all currently undisclosed (Anthropic; Finimize notes the baseline is undisclosed). An escalator that grows with valuation works like an option the board can exercise against shareholders. Some investors will price that as a cost. Others will see it as reducing political risk. Rival labs, OpenAI in particular, may face questions about whether they will match it.

For the field and for policy. The debate over AI and jobs has mostly been about forecasts and taxes. This is a private, voluntary mechanism for "sharing the upside", and it will be watched as a test case. If it works, governments might point to it as a model, or treat it as a reason not to legislate. If it stalls, critics will point to it as pre-IPO public relations. One second-order effect: because the fund's equity is barred from lobbying, a pot of this size can't be turned directly into political influence. That rule would matter a great deal for a vehicle that could be worth tens of billions of dollars.

What to be skeptical of

It's a pledge, not a transfer. The 5% depends on a stockholder vote whose date and mechanics are unpublished. The escalator is discretionary, and its baseline is unknown.

Company control. The fund is wholly owned by Anthropic. Unlike a registered charity, it doesn't automatically face outside oversight of its spending. Anthropic promises governance details "in the coming months", and until then readers can't judge how independent it will be. A further open question, in our view: a 5%-plus block of voting stock is a source of influence in itself, and Anthropic has not yet explained how the fund's shares will be voted.

Founder control in the background. Corporate-law researcher Marco Mari argues that the reported Founder LLC and LTBT arrangements would leave public investors with no independent board election rights. Insulation that protects a mission, he writes, can also shield the people interpreting it from challenge. The fund adds one more pot of value that this structure will direct.

The valuation is doing the heavy lifting. "$48 billion" is 5% of a private round price. Public markets could value the company quite differently, and the fund's real-world purchasing power depends on when and how shares are sold.

Timing. The announcement lands weeks before an expected listing and days after a public call to pause the IPO. Tech Insider, pointing to reports of pre-IPO investor pitches days earlier, called the timing "unlikely to be coincidental". That doesn't make the pledge insincere, but it does make independent checks more important.

Different philosophies. Elon Musk, who runs a competing AI company, replied on X: "I recommend giving the shares directly to the public", adding that this is what he will be doing. That is a real design alternative: broad ownership instead of a managed fund. It doesn't carry the same weight as criticism from a neutral party.

What to watch next

1. The stockholder vote. When it happens, what share of votes approves it, and whether the terms change.
2. The baseline and milestone definitions. Is the baseline the $965B Series H price, the IPO price, or something else? Are the steps additive or compounding? Anthropic promised these "in the coming months".
3. Governance of the fund. Who sits on its board, whether outside parties have any say, and how the fund votes its shares.
4. The public S-1. Whether the prospectus, once public, describes the fund, its share source and its dilution effect. Also whether Anthropic answers the NYC Comptroller's letter.
5. The first public benefit report. Anthropic says it will report on progress against its mission. Look for spending timelines and first recipients.
6. Responses from rival labs. Whether OpenAI, Google or others announce comparable equity-based commitments.

For more on how Anthropic describes its own models' risks, which is the backdrop to its governance choices, see our explainer on Anthropic's "persistence" findings. For a different idea of where AI safety should sit, see Nadella's "emergency brake".

How this was made: researched and written by an AI model (Claude) from the primary sources listed above, then checked claim-by-claim against those sources in a separate AI fact-check pass. Spotted an error? Email [email protected] and we correct it publicly. Our process.