A public company, controlled by seven people
When Anthropic filed its S-1 last week, the world saw breathtaking numbers: $16.3 billion in revenue in the first half, a growth curve that had investors pricing the company at up to $2 trillion. But what went unnoticed under the spotlights was the governance architecture that transforms Anthropic into something radically different from any other publicly traded tech company — perhaps into something even more troubling for those who believe in free-market capitalism.
The company created not one, but five classes of stock. Class A, with up to 300 billion shares, will be publicly traded. Class E, with up to 30 billion, has minimal economic rights. Class N, with up to 100 billion, has no voting rights. Class T, a single share, controls the board election. And Class F — by far the most extraordinary — is a single share with zero economic rights, holding 50.1% of total voting power in all ordinary shareholder deliberations. That share is held by an LLC formed by the seven co-founders.
To put it in perspective: each of the seven co-founders owns about 2% of the company's economic shares — and has pledged to give away 80% of their wealth. But the Founder LLC, controlled by the majority of the seven, makes decisions on 50.1% of corporate votes. In other words: seven AI engineers can approve or reject any merger proposal, any acquisition offer, any board swap — even though they hold an infinitesimal fraction of the economic capital. And this holds as long as at least three of the seven founders retain a minimum stake.
When only two or fewer founders (or their successors) remain, Class F will begin a gradual sunset process.
Inspired by Palantir, borrowed from Zuckerberg and Musk
The structure isn't original — but it's the most extreme ever seen in an AI company.
Class F is modeled on Palantir's dual-class structure, where Alex Karp and Peter Thiel maintain disproportionate control. It's inspired by Meta's multiplicative voting shares (Zuckerberg has over 60% of votes). And it echoes Tesla/Musk's share class, where the CEO holds a massive slice of decision-making power regardless of shareholding.
But what differentiates Anthropic is the collective model. No other company of comparable scale has handed 50.1% of votes to a founding board. In the others, it's one or two individuals. In Anthropic, it's a committee of seven people — which makes even harder any attempt at external challenge, since you'd need to convince all seven, not just a single CEO.
The central contradiction
Anthropic has long positioned itself as the safety-conscious AI lab. In its very S-1, the company warned about "catastrophic or existential risks to humanity." But what the governance structure reveals is that the very idea of "responsible control" is concentrated in hands that don't need to answer to anyone — not even shareholders who paid billions for the shares.
The company acknowledged explicitly, in its governance documentation, that this structure could lead to "decisions that may conflict with short-, medium- or long-term financial interests and business performance." In other words: the seven founders can make choices that cost money — and shareholders will have to swallow it.
What this means for the AI sector
Anthropic's rival, OpenAI, is also seeking a founder-control structure — but is still in private fundraising phase. If Anthropic lists in November successfully, OpenAI will have pressure to adopt similar structures. And if OpenAI does the same, we'll be facing a scenario where the world's two largest AI labs are effectively companies controlled by founding committees that don't answer to shareholders.
It's the end of free-market capitalism for the AI sector — at least at the highest levels. And the question isn't whether this structure is good or bad, but whether the artificial intelligence industry has already become so concentrated that corporate democracy has become an irrelevant detail.
Sources: The Verge, Fortune, CNBC
✓ Independent sources cross-checked and verified before publishing
