- The US administration’s decision to block foreign nationals from accessing Anthropic’s most advanced models — prompting the company to disable them for everyone — marks the first time Washington has directly intervened in an AI lab’s model availability, not just its chips.
- Goldman Sachs partner Bobby Molavi framed frontier AI as “state-supervised strategic infrastructure,” warning that model availability is now an operational risk for both AI providers and enterprise users, with implications for IPO valuations and stock multiples across the Magnificent Seven.
- Rosenblatt Securities analyst Barton Crockett called it an “obvious disruptive risk to the entire AI growth story,” arguing that political risk is now “a permanent line in every AI valuation” — discounted into IPOs and shaved off multiples for anything touching AI sovereignty.
- China is positioning itself as the beneficiary: Zhipu unveiled its most advanced open-source model within hours of the Anthropic curbs, its stock surged 33%, and DeepSeek has made a permanent 75% price cut on its flagship model — accelerating a global bifurcation in AI access.
What Happened?
After the US government ordered restrictions on foreign access to Anthropic’s most capable models — citing national security concerns — Anthropic disabled the models for all users globally rather than implement a geographic filter. The company is in talks with Washington to restore access, but the episode has already redrawn the risk map for AI investors. Anthropic’s pre-IPO proxy contract on Hyperliquid slipped roughly 3.7% to around $1,627, a small move in absolute terms but meaningful given the company’s reported ~$965 billion valuation ahead of a planned IPO. Notably, Anthropic already sits on the Department of Defense’s supply-chain risk list after refusing to let Claude power mass surveillance or fully autonomous weapons.
Why It Matters?
For years, AI regulatory risk was about chips — export controls on Nvidia GPUs to China, for example. This episode extends the reach of government intervention to the models themselves, creating a new category of “platform risk” that runs not just through pure-play AI firms but through every Magnificent Seven company with stakes in or contracts with frontier AI labs. Goldman’s Molavi asked the core question for equity investors: will AI model providers become “hybrid consumer, enterprise and state-funded defense contractors”? The European sovereignty trade is already responding — France is replacing Palantir with domestic tools, Mistral AI’s valuation is doubling in new funding talks, and OVH Groupe shares hit their highest since 2023.
What’s Next?
Anthropic’s negotiations with the White House over restoring model access will be closely watched — if the jailbreak concerns are as narrow as the company insists, access could return within weeks. But even a swift resolution won’t erase the precedent: government can flip a kill switch on frontier model availability. That risk becomes fully priceable once Anthropic and OpenAI go public, transforming what is now a theoretical tail risk into a directly tradable discount. Meanwhile, China’s rapid open-source response — Zhipu, DeepSeek, and others filling the gap left by US restrictions — suggests the bifurcation of global AI access is accelerating faster than most Western investors had modeled.
Source: Bloomberg
