Anthropic Just Made Shareholders Disappear

Anthropic just made a legal argument that could leave thousands of investors holding nothing.

The company updated its support page on May 12 with language that securities lawyers describe as unusually aggressive. Any transfer of Anthropic stock without board approval, the policy now reads, is "void and will not be recognized on our books and records." Not voidable. Void. Anthropic Support Page

The distinction matters enormously. A voidable contract is one that a court can choose to unwind after the fact. A void contract is one that never legally existed in the first place. If Anthropic's position holds, anyone who bought Anthropic shares through an unauthorized channel did not acquire an ownership stake. They acquired a lawsuit.

"Investors should pay very close attention to that word," said a securities attorney with knowledge of secondary market disputes who requested anonymity to speak candidly. "Anthropic is not trying to unwind these transactions. They're trying to make them disappear entirely. That's a fundamentally different legal posture, and it has real consequences for how courts and regulators will respond."

The company explicitly named eight firms as unauthorized intermediaries: Open Door Partners, Unicorns Exchange, Pachamama, Lionheart Ventures, Hiive, Forge, Sydecar, and Upmarket. Decrypt Some of these platforms have operated for years as legitimate marketplaces for private tech shares. Two of them, Forge Global and Hiive, are regulated secondary marketplaces that host auctions, provide custody services, and facilitate valuations for private company equity. Anthropic's decision to include them in the same category alongside obvious scams sends a message that the company intends to treat the entire unauthorized secondary market as compromised, regardless of how it was structured.

The policy update came on the same day OpenAI published nearly identical language on its own equity policy page. OpenAI Policy OpenAI's version declares any transfer without written consent "void," uses the same taxonomy of prohibited structures, including SPVs, tokenized interests, and forward contracts, and was published on May 15, three days after Anthropic's. Both companies raised approximately $5 billion in their most recent funding rounds at valuations that place them among the most valuable private companies in history. The timing and legal parallelism are unlikely coincidental. Industry sources suggest the two companies coordinated their language, though neither confirmed that directly.

The backdrop is a secondary market that has expanded far faster than the legal infrastructure to govern it. Over the past five years, a dense ecosystem of exchanges, SPVs, and fractional ownership platforms emerged to serve employees and early investors at AI labs who wanted to monetize paper wealth without waiting for a public offering. These platforms processed hundreds of millions of dollars in transactions annually. The companies largely tolerated them, neither endorsing nor explicitly prohibiting the trades happening in their names.

That tolerance is ending. OpenAI's $6.6 billion employee tender offer, completed this year as a board-authorized process with every transfer individually approved, stands in direct contrast. Decrypt It is the legal line Anthropic and OpenAI are drawing: company-controlled tender offers are legitimate; everything else is void.

The revenue trajectory explains part of the urgency. Anthropic's annualized revenue tripled from $9 billion at the end of 2025 to $30 billion by April 2026, driven largely by Claude Code adoption among enterprise software developers. Decrypt At those valuations, the gap between what early investors paid and what unauthorized secondary buyers have been willing to pay is enormous, and the incentive to arbitrage that gap through SPVs, tokenized wrappers, and secondary platforms has been correspondingly large. Amazon has committed up to $25 billion in Anthropic, a investment that has no obvious near-term liquidity path and therefore creates pressure for alternative mechanisms to unlock value. The crackdown is an attempt to close that gap by making unauthorized transfers legally worthless.

The secondary market appears to have reacted quickly. The Anthropic token on PreStocks, a Solana-based platform that sold synthetic Anthropic equity exposure through an SPV-like structure, dropped from $1,400 to $900 after the announcement, a 36 percent loss in hours. Decrypt Forge Global and Hiive did not respond to requests for comment by publication.

What happens to people who already bought through these channels is genuinely unclear. Anthropic has not announced enforcement actions, and the company's support page reads as a statement of policy rather than a notice of litigation. But void is a strong word in securities law. If Anthropic's board never approved a transfer, the company may have grounds to refuse to recognize the new holder on its books and records, regardless of what the investor paid or what a secondary exchange shows as the current price.

The deeper question is what this means for the premise underlying the entire private equity secondary market. For two decades, the logic of that market has been that economic exposure to high-growth private companies could be packaged, resold, and democratized through platforms that aggregated capital from investors who could not or preferred not to hold shares directly. That premise depended on a legal assumption: that whoever holds the shares, even if the transfer was technically restricted, has a real economic interest that courts would recognize and protect. Anthropic and OpenAI, acting in coordination or in parallel, have just rejected that assumption for their own cap tables.

The precedent, if it holds, extends well beyond these two companies. Any AI lab with similar transfer restrictions now has a legal template for invalidating secondary market activity that it did not authorize. The investors who bought through these platforms (many of them wealthy individuals and family offices who believed they were participating in the AI economy in a meaningful way) are the ones absorbing the risk.

This is a property-rights earthquake disguised as a support page update. Anthropic has decided it wants complete control over who owns its equity, and it has chosen the legal theory most favorable to that goal. Whether courts agree is a separate question. For now, anyone holding Anthropic or OpenAI shares through an unauthorized channel should read that word carefully: void.