The Public Gets the Upside; the Founders Keep the Votes

Anthropic explained its new public benefit fund in the least philanthropic way available ^[1].

It is common practice for tech companies to offer escalating compensation awards to their executives. We are instead directing that equity to the public benefit. — Anthropic

That sentence is the whole instrument in miniature. The pledge is not charity bolted onto a listing; it is the thing that happens instead of bigger pay packages. Framed that way, it becomes a cap-table decision: a choice about who owns what, not who gets thanked. Read the fine print and the fund stops looking like a gift and starts looking like a structure. It is a wholly owned LLC, not an independent trust. No share ever passes into a citizen's hands, because Anthropic controls the vehicle outright. Its stated beneficiaries are narrower than "the public."

Given the pace of AI progress, bolder approaches may be needed. The fund will have the versatility to pursue them. — Anthropic

That equity is barred from lobbying, and it scales with the founders' wealth rather than the public's need. The stake seeds at 5% and adds two percentage points for every 50% the company's value rises above a baseline ^[1]. Ten months of beats lead to it, and each pairs a political exposure with a market answer. The sequence opens in December, with Amodei telling 60 Minutes he was "deeply uncomfortable" with these decisions resting with a few companies ^[2]. In February, Anthropic made its first known political donation: $20 million to candidates who favor AI safety regulation, against the $125 million OpenAI and Andreessen Horowitz are spending on the other side. That same month Amodei warned AI could mint "personal fortunes in the trillions" almost by accident ^[3][4]. In May it sued the Defense Department over what it called retaliation for refusing the military unrestricted model access ^[5]. In June it launched Claude Corps, a $150 million program embedding a thousand AI fellows in nonprofits, plus a separate $200 million to study and support workers displaced by AI, the same month it filed confidentially to go public ^[6]. In September the prospectus landed with both tracks in one document: a target near $2 trillion, and a Founder LLC holding a single Class F share carrying 50.1% of the votes, while the co-founders pledged 80% of their personal equity to charity ^[7]. Then, three weeks out from the listing, the run tightened to a daily rhythm ^[8][1]. Investor day on October 14, revenue against doubts about whether the capex pays back ^[9]. October 16, the company's own survey finding most of its 1,334 employees expect AI to handle at least 80% of their work by 2030 ^[10]. October 17, its co-founder telling the IMF the number could be $44 trillion in GDP ^[11], and its research adding janitors and gardeners to the jobs robotics now threatens ^[12]. The drumbeat ran four days, investor day to the fund; the research that maps the displacement and the vehicle that promises to share the spoils landed within two days of each other. What the fund disperses, against what never moves, is the tell. The economics go into a company-controlled vehicle aimed at displaced workers. The vote goes nowhere. The Class F share keeps 50.1% with the founders ^[7], and the Long-Term Benefit Trust, Ben Bernanke among its members, holds the power to elect a majority of the board through its own non-economic shares ^[13]. Elon Musk's reply was plain: "I recommend giving the shares directly to the public." ^[1] Golden Door's Vijar Kohli read the same design from the other direction: a wholly owned fund keeps those shares under Anthropic's control and pre-emptively dilutes outside voting power rather than dispersing it ^[1]. The political arithmetic explains the shape. Two weeks before the fund, Alexandria Ocasio-Cortez told a town hall what she says the people getting rich from AI have in mind for everyone else ^[14]: "…keep all of the spoils of those riches for themselves." The fund's targeting sentence names the same population the charge names. Morgan Stanley has documented that what protects high-income households from AI disruption is equity wealth, while Apollo estimates 5.8 million exposed workers lose roughly $28 billion a year in wage growth; 75 data-center projects were blocked or delayed this year and New York enacted the first statewide moratorium on large data centers in July ^[15]. Set all that beside Alex Karp's September proposal that nationalization, a government stake, is the only real cap on AI liability ^[16], and they are two versions of the same ground: one hands the public up to a fifth of the upside inside a company-owned box, the other hands the state half. The stake is real money, around $100 billion at launch and up to $400 billion at a fifth of a $2 trillion company ^[1], which is why the design, not the sincerity, is the part worth reading. The design is already table stakes. The day after the fund, Musk pledged one SpaceX share, roughly $146 of private, illiquid paper, to every American child, "the first of many" ^[17]. OpenAI's nonprofit already holds 26% of the company ahead of its own listing ^[18][19]. Every entrant is now offering the same trade: the public's upside, the founders' votes.