Anthropic's Public Benefit Fund and the Debate Over Who Shares AI's Wealth

On October 18, 2026, Anthropic announced the Anthropic Public Benefit Fund and said it would place 5% of its shares in it, with more to follow if its valuation keeps rising. Counting earlier pledges from its co-founders and staff, roughly a fifth of Anthropic’s equity is now set aside for the public good. At the $1.5 trillion to $2 trillion valuation rumored for a future IPO, the first 5% alone would be worth $75 billion to $100 billion. Anthropic frames the move as preparation for an economy in which AI produces enormous value while unsettling many workers; critics see public relations, a governance problem, or both. The debate also reopens an older question: if AI generates a windfall, how should it be shared, and is universal basic income (UBI) the right tool?

How the fund is built

Anthropic laid out the structure in a post titled “Investing in the public benefit.” The main terms:

• Initial grant: 5% of company shares, committed once stockholders approve.
• Top-ups: the board may add another 2% each time the company’s value climbs 50% above a set baseline.
• Legal form: a limited liability company (LLC) wholly owned by Anthropic, rather than a 501(c)(3) charity or a Public Benefit Corporation. The choice gives the company the most room to decide how the money moves.
• Uses: large donations to established nonprofits and direct cash payments to specific groups of people. Broad dividends to every citizen are not part of the initial plan.
• Limits: none of the LLC’s equity may fund political lobbying or campaign donations, in the United States or anywhere else.

The company also committed $200 million to studying and responding to AI-driven disruption in the labor market.

The “roughly 20%” figure combines four separate commitments:

Co-founder Jack Clark wrote on X that setting aside about 20% of the company spreads AI’s benefits widely while building a financial buffer against the technology’s risks and against job displacement.

How large it could become

At the rumored IPO valuation, the initial 5% would rival three familiar yardsticks:

That could make it the largest pool of philanthropic capital ever assembled.

The baseline for top-ups has not been published, so any projection is illustrative. Starting from a $2 trillion IPO and assuming the value rises in 50% steps, the math looks like this:

Figures near the top of that range assume AI automates much of high-value knowledge work.

Why Anthropic wants a cushion

In its announcement, Anthropic warned that AI could create vast economic value while seriously destabilizing the labor market. It also conceded that frontier labs, the companies building the most advanced models, do not yet know how large that disruption will be or what path it will take.

The warning echoes Anthropic CEO Dario Amodei. In May 2025 he said AI could wipe out half of entry-level white-collar jobs within one to five years and push unemployment to 10% to 20%. He floated a 3% tax on tokens, the units of text that AI models process, to pay for social safety nets.

Capability gains feed the unease. A claim circulated that an internal OpenAI model, with compute equal to about three hours of ChatGPT Pro thinking, had solved roughly 20% of the top open problems in mathematics, or 90 of the top 500 by a follow-up tally, alarming mathematicians about their own roles. Similar abilities appear to be emerging in physics and may spread to other white-collar fields.

Not everyone accepts the gloomy forecast. David Sacks, the former U.S. AI czar, argues that forecasters should be judged on their track records, and he points to employment data showing that white-collar technical jobs have grown rather than shrunk. When coding agents such as Claude Code arrived the year before, many expected senior L5 and L6 software engineering roles to vanish quickly. Instead, companies have hired more engineers to supervise and coordinate growing fleets of AI agents, software that carries out multi-step tasks on its own.

Anthropic’s June 2026 Anthropic Economic Index report adds nuance: in a survey of Claude users and non-users, people who automate the most of their work with Claude were the most optimistic about job security, pay, and the meaning of their work.

The backlash: who defines the public good?

Reaction on social media and in financial commentary was largely skeptical.

• A pre-IPO image play: one critique casts the pledge as calculated public relations aimed at retail investors ahead of an IPO.
• A bailout hedge: one commentator argued that giving the public a financial stake in the company would build political support for government help if the company ever ran into serious trouble.
• No guarantee of impact: David Sacks said that parking 20% of a company’s equity in an LLC does not by itself deliver public benefit, noting that deploying philanthropic capital well at large scale is still an unsolved problem.
• Whose public: Palmer Luckey, co-founder of Anduril Industries, asked who inside Anthropic decides what counts as the public good. He compared the plan to “USAID 2.0,” suggesting the money could flow abroad instead of to Americans.
• The EA angle: former Uber CEO Travis Kalanick called the fund an “EA infinite money glitch,” a jab at Effective Altruism, the movement Anthropic is ideologically close to.

Some concerns follow from the structure itself. Because founders and major shareholders govern Anthropic, the fund’s priorities are likely to reflect leadership’s preferences more than any broad public consensus. Letting the company alone choose which “specific populations” receive cash also leaves accountability murky.

On the other side, committing about a fifth of a private company’s equity to public benefit is unusually generous by modern corporate standards. And whatever one thinks of Effective Altruism, the risk of serious harm from AI systems developed without adequate limits seems hard to dismiss entirely.

Three ways to hand out AI wealth

Anthropic’s fund is one of three models now on the table for spreading the gains from AI.

▲ Three ways to distribute AI wealth

In a pointed reply to Anthropic, Elon Musk argued that equity should go straight to the public rather than through intermediary foundations. On October 19 he pledged to give one SpaceX share to every child in America in 2027, calling it the first of several distributions. With roughly 73 million children in the country and an illustrative price of $135 per share, that comes to about $10 billion. A business commentator argued similarly that charities lose money to executive pay and bloated boards, while capital kept in productive companies creates more jobs and tax revenue.

OpenAI floated its Public Wealth Fund in April 2026 as a way to pay citizens recurring dividends from AI-driven productivity. Its restructuring in October 2025 had already left the OpenAI Foundation holding equity worth roughly $130 billion to support public-benefit work.

Each model still has to answer the same hard questions: who defines public benefit, and who qualifies to receive it?

What UBI means, and what one trial found

UBI rests on three ideas:

By that definition, Anthropic’s fund is not UBI but an equity endowment with performance-linked top-ups.

Funding basic income from AI’s gains is an idea about a decade old. Sam Altman argued in 2016 that basic income is not socialism. In 2021 he proposed an American Equity Fund that would tax large companies 2.5% a year on market value above a certain threshold, along with 2.5% on privately held land, and pay about $13,500 a year to each of 250 million adult Americans.

The most concrete evidence comes from a guaranteed income study that Altman funded:

Musk has a bolder vision he calls “universal high income.” In his telling, AI and robots handle all physical and intellectual labor, poverty disappears, and nobody needs to save money, leaving people free to spend time with family, make art, or invent. The trial results suggest unconditional cash does not automatically unleash a wave of creativity; participants mostly worked a little less. Whether UBI or equity distributions would work at scale remains unproven.

Work, purpose, and the agent manager

Beneath the money lies a question about meaning. For millennia, people have found purpose in invention, discovery, literature, and art. Earlier technologies, from industrial machines to computers, extended what people could do. Frontier AI automates thinking itself, while robots take on physical work, joining a digital mind to a mechanical body.

Two pictures of what follows compete:

• The bleak view: full automation leaves people passive, monitored, and treated like children by superintelligent systems.
• The hopeful view: AI takes over clerical chores such as spreadsheet entry, email sorting, messaging, and information gathering, and people become managers who direct fleets of AI agents and spend their time on strategy, creativity, and decisions.

▲ A person directing a team of AI agents

The hopeful view expects real friction but sees leaning into agents as the clearest way to keep human purpose intact. Either way, separating work from survival would unsettle education, careers, and the status that degrees, certifications, and corporate ladders have long conferred. Unease is not doom-mongering when even the companies building frontier AI say they cannot see where this ends.

What to watch and what to do

Some facts are clear: Anthropic has earmarked about 20% of its equity for public benefit, the fund is an LLC whose board decides who gets money, and none of its equity can pay for political lobbying or campaigns. Whether the pledge is real altruism or pre-IPO positioning, which distribution model works best, and whether unconditional money frees people or mainly leads them to work less are judgments the evidence has not yet settled.

Practical steps for readers:

1. Read the structure of any corporate pledge. Check whether the money sits in a charity, a Public Benefit Corporation, or a flexible LLC, and whether payouts depend on board approval or valuation triggers.
2. Keep UBI and equity gifts separate. Recurring payments to everyone differ from one-time or vesting equity donations.
3. Track AI labor data. Indexes such as the Anthropic Economic Index show which white-collar roles face automation pressure and which are absorbing AI as a collaborator.
4. Shift your work toward delegation. Hand spreadsheets, email triage, and first-pass research to AI agents, and build the skills to assign and review their output.
5. Watch frontier results in math and physics. They signal how quickly expert work elsewhere may change.