Last Friday, the White House ordered Anthropic to block export of its AI models Fable and Mythos to anyone outside the U.S., including foreign nationals inside the country. Anthropic pulled both models within 90 minutes, leaving the tech world stunned by the speed of the restriction in early week news.

The move follows two triggers: a South Korean telecom—SK Telecom—received Mythos through a limited‑partner program, raising U.S. concerns over possible Chinese ties, and Amazon researchers reportedly bypassed Fable 5’s safeguards. Anthropic disputed the jailbreak claim, calling the issue narrow and patched in the same week as the ban was issued again.

Historically, U.S. export controls have struggled to curb cyber tech spread. The 1990s PGP controversy and later Wassenaar arrangements illustrate how encryption and spyware export limits often fail, especially when countries like Israel or Italy sidestep rules. These precedents shape the current debate over AI containment.

Anthropic now faces a choice: lift the ban and risk losing market access, or seek U.S. approval for every foreign customer, a compliance burden that could dent profits. The outcome will set a precedent for how other AI labs navigate export rules and shape the industry’s regulatory landscape in 2025.