Anthropic recently doused the enthusiasm of investors who wanted to get on board early.

This unicorn in the large model space suddenly tightened restrictions on over-the-counter stock transactions, clearly warning that any share transfer not approved by the board will not be recognized by the company. Even more "hardline" was the official support page directly naming multiple secondary markets and private platforms, stating that the Anthropic stocks or related rights circulating on them are unauthorized, and such transactions may be invalid, with shareholder registers not being updated.

The news immediately caused tension within investor communities. Bloomberg reported that in a WhatsApp group consisting of hundreds of family office members, someone bluntly asked: "Are we doomed?"

Beneath this is an "expectation mismatch" in the frenzy of AI valuations.  

As one of the most closely watched companies in the current competition for large models, Anthropic has already been labeled as "must-have" by the market, even though it hasn't gone public yet. However, the problem is: ordinary people can't buy shares of unlisted companies. As a result, various "workaround solutions" have emerged—special purpose vehicles (SPVs), secondary market share transfers, forward contracts, nominee structures, and even various financial products that claim to allow "indirect holding" of equity in hot AI companies.

These transactions satisfy a fear of missing out: by the time it actually goes public, it might be too expensive; if you don't invest now, you might miss your chance in the future.

But from the company's perspective, the logic is entirely different.  

Especially during periods when valuations are rapidly rising and the potential for going public is increasing, companies like Anthropic need to strictly control their shareholder structure, transfer paths, and the pace of information disclosure. If the external market overhypes various "shadow rights," it could not only interfere with corporate governance but also bring compliance risks and reputational backlash.

Therefore, this reaffirmation of rules is essentially a "boundary clarification": you may think you've bought an early ticket to the AI unicorn, but the company may not recognize the validity of that ticket.

This is indeed harsh for investors, because the appeal of many pre-IPO transactions lies precisely in the imagination of "locking in the gains from the listing." Once the company clearly states that unapproved rights are invalid, those who bought through complex structures may suddenly realize: what they hold is not a stock, but a contract with a huge room for interpretation and unclear path to redemption.

This also explains why investor groups are so anxious. The higher the valuation of AI companies, the more people want to get in early; the more people try to get in through alternative routes, the more the company needs to tighten its boundaries. The result is: the hottest assets have become the most opaque and hardest to verify ones.

Looking at the industry as a whole, this incident exposes an awkward reality: the market has already priced top AI companies as the "next internet," but their business models, revenue paths, and cost structures haven't really been proven yet. Computing power costs are pouring in like water, income fluctuations are hard to predict, and the industry shakeout is far from over. In the context of high valuations and lack of transparency, it's easy to form a collective illusion where everyone is afraid of missing out, but no one fully understands the situation.

Even more worrying is that when the wealth glow of AI unicorns spills over from the primary market to attract ordinary investors, the information asymmetry gap is actually widening. Ordinary people often bear the highest risk without getting the most basic protections. Even if the equity transfer itself is valid, the assets they bet on remain full of uncertainty.