The conflict between Anthropic and the US government over the suspension of Fable 5 and Mythos 5 is getting sharper by the day. What started as a dry export-control order now comes with a concrete suspicion, public rebuttals, and an appearance from the Pentagon.
The new accusation: Chinese access
According to a June 14 report by Semafor, the order was issued partly over suspicions that a China-linked group may have accessed Anthropic’s new model. David Sacks, an adviser to President Trump, described on X how the government learned about the security issues: there had been a warning that Fable 5 could be jailbroken — and when CEO Dario Amodei was notified, the company allegedly did nothing about it.
The Pentagon’s chief information officer, Kirsten Davies, publicly backed the national-security line on June 13: “Some things are simply more important than revenue cycles, clickbait, and pre-IPO valuation.”
Anthropic’s counter-account
Anthropic sees it differently. In its blog post, the company says the government provided only “verbal evidence of a potential narrow, non-universal jailbreak.” And it adds that it disagrees a narrow potential jailbreak should be cause for recalling a commercial model deployed to hundreds of millions of people.
The net effect of the order: Anthropic has to abruptly disable Fable 5 and Mythos 5 for all customers to stay compliant. Every other Claude model remains available. Anthropic calls the whole thing a “misunderstanding” and says it’s working to restore access as soon as possible.
Who gets hit?
The order locks out every foreign national — inside and outside the US, including foreign-national Anthropic employees. Several key people, like co-founder Chris Olah, AI researcher Andrej Karpathy and philosopher Amanda Askell, were born outside the US. Whether they themselves lose access is unclear. Companies like S&P that have integrated Claude into their databases, and workers on H-1B visas, are affected too.
My take
Two logics are colliding here. The government argues national security; Anthropic argues proportionality. Neither side can be verified independently — and that’s exactly the problem. If “foreign national” is the criterion you use to secure a model, it’s barely enforceable in practice and easy to bypass. What’s left is a precedent: AI models are becoming export goods with security clearance. And the rift between Anthropic and Washington — already simmering over a lawsuit about military use — just gained another uncomfortable front.
Note: Several details rest on media reports and statements from individual parties that cannot be independently verified.
Sources: