Anthropic Export Controls Recast AI Safety as Nationality Test

When Anthropic was directed to suspend access to its Fable and Mythos systems for foreign nationals, the scope of the restriction landed with particular force. According to Armin Ronacher writing on his blog, "the directive, as Anthropic describes it, applies to foreign nationals whether they are inside or outside the United States, including foreign national Anthropic employees." The directive does not distinguish between hostile governments and allied democracies. Nationality is the test.

Ronacher frames the underlying shift precisely: "We moved from 'do not sell this model to hostile governments' to nationality itself being the defining boundary." That compression matters. The earlier rule was defensible on security grounds. The new one implies something broader: that only Americans can be trusted with the most capable AI, regardless of employer or location.

The piece lands hardest on what this reveals about AI safety rhetoric. Ronacher observes that "a lot of AI safety discourse presents itself as universal: humanity, catastrophic risk, safeguards, responsible deployment" while carrying an undertone of national security interest. "The foundation is that the US has moral superiority and others are not to be trusted," he writes. Safety, in this reading, was always partly a national security argument wearing a universal costume.

For Europe, Ronacher sees a structural dependency that is difficult to ignore: "We depend on American cloud providers, operating systems, developer platforms and now AI models." Self-inflicted wounds compound it. "We built and maintained fragmented markets and then pretended we had a single one," he writes, and the result is a feedback loop he calls a "dangerous death spiral already. Talent leaves because the ecosystem is weak and the ecosystem stays weak because talent leaves."

His answer is international cooperation and open-source development, which he describes as "one of the few paths we have that does not naturally lead to total concentration of power." What the piece does not give you is a concrete policy mechanism or a named institution that could enforce that cooperation. It is a diagnosis more than a prescription. But the underlying claim that AI is "quickly becoming another instrument of militarization and national rivalry, when it could be one of the most powerful tools for cooperation we have" gets harder to dismiss the more nationality-based access restrictions accumulate.