Anthropic spent its weekend in a standoff with the Trump administration over its latest AI models, Claude Mythos 5 and Fable 5, after receiving a US export control directive at 5:21 PM on Friday. The order required the company to suspend access to the models by any foreign national, including its own overseas employees. In response, Anthropic chose to disable its newly launched products and send executives to Washington, DC, to negotiate. CEO Dario Amodei personally joined calls with US Treasury Secretary Scott Bessent, Commerce Secretary Howard Lutnick, and National Cyber Director Sean Cairncross. The government's action followed a report that a potential jailbreak—described as narrow and non-universal—had been shared with authorities by an unnamed entity. Anthropic stated that the capability demonstrated in the report was already available in other models, including OpenAI's GPT-5.5. The company emphasized that the vulnerability was not unique to Fable 5, which had been marketed as safe for general use due to additional safeguards. Earlier concerns reportedly stemmed from fears that a China-linked group had accessed the technology, linked to a global telecommunications firm initially approved for early access to Mythos Preview. Anthropic revoked that access when the government raised alarms. David Sacks, the US government's AI envoy, referenced a "highly credible trusted partner of both Anthropic and the USG" who identified the jailbreak during testing. Some reports suggest Amazon CEO Andy Jassy alerted officials after Amazon researchers conducted red-team exercises on Fable 5. The Trump administration, previously hands-off on AI safety, has shifted tone amid fears of losing the AI race to China. Cybersecurity leaders now warn that halting Mythos 5 and Fable 5 could inadvertently boost China's position in AI development. The incident has accelerated industry moves to sign backup contracts with non-US firms and deploy open-weight models on alternative hardware.
Anthropic spent months warning that its own AI was too dangerous to release, then launched a version it claimed was safe—until the government stepped in. The company's push to commercialize Fable 5 despite its own cautionary narrative weakens its credibility. If the jailbreak in question exists in other models like GPT-5.5, as Anthropic admits, then the panic over its tech being compromised is more about optics than unique risk. Nigerian AI startups watching this should note: regulatory survival now depends as much on political alignment as on technical safety.
Editorial note: AI-assisted opinion, not established fact. Full disclaimer →
