U.S. regulators blocked foreign users from Anthropic PBC’s Claude models over security worries, forcing the company to shut access for everyone. The move marks the first time a government has stepped in on AI model distribution, raising a new political risk that could hit earnings of firms that rely on global model sales across industries.

Anthropic’s action follows a U.S. administration order that barred foreign nationals from accessing its most advanced models, a first for a leading AI lab. The shutdown signals that model availability is now an operational risk, especially if sanctions or security reviews appear suddenly, a concern echoed by Goldman Sachs partner Bobby Molavi who called frontier AI “state‑supervised strategic infrastructure.”

The episode reshapes AI valuation logic. Investors now weigh the possibility that model providers could become hybrid consumer‑enterprise‑state funded defense contractors, a shift that could dilute earnings projections for tech firms and chip makers whose revenues depend on unrestricted model access. The thin over‑the‑counter market for Anthropic’s pre‑IPO stock slid 3.7% to roughly $1,627, a negligible move against the sector’s trillions‑level cap.

With Anthropic in talks to lift the curbs and Beijing’s Zhipu rushing out a new open‑source model, governments now dictate who can own the best AI tools. The precedent forces the market to discount political risk into IPO pricing and earnings multiples, meaning any future AI growth story will carry a sovereign‑risk surcharge that investors must price in today.