At 5:21 p.m. ET on Friday, Anthropic was hit with a U.S. export‑control order that demanded an immediate shutdown of foreign‑national access to its latest models, Claude Mythos 5 and Fable 5. The directive, issued by the Commerce Department, applied to anyone inside or outside the United States, including Anthropic’s own foreign‑national employees. Faced with a 90‑minute deadline, the company had to decide whether to comply or risk a broader set of sanctions.
Anthropic’s leadership responded within minutes. CEO Dario Amodei joined a conference call with Treasury Secretary Scott Bessent, Commerce Secretary Howard Lutnick, and National Cyber Director Sean Cairncross. The discussion quickly moved from a phone call to an on‑the‑ground effort, with Anthropic’s head of safeguards Dave Orr, red‑team lead Logan Graham, and cybersecurity researcher Nicholas Carlini boarding a flight to Washington, D.C., later that day.
The government’s concern stemmed from a report suggesting that a “jailbreak” technique could bypass the safety guards built into Fable 5. Anthropic described the issue as a narrow, non‑universal vulnerability that had been shared with the administration by an unnamed, trusted partner. The company also noted that similar capabilities existed in other models, including OpenAI’s GPT‑5.5, and that the reported exploit was not unique to its own technology.
Sources familiar with the negotiations said the administration’s alarm was heightened by rumors that a China‑linked telecommunications firm had accessed the technology weeks earlier. While some reports pointed to Amazon’s red‑team research as the trigger for the government’s action, the exact chain of events remains unclear. What is clear is that Anthropic revoked access to the implicated parties as soon as the concerns were raised.
Anthropic’s internal documents show that Mythos 5 was initially rolled out to a limited set of government agencies and corporate partners, while Fable 5, equipped with additional safeguards, was marketed as safe for broader use. The sudden export‑control order forced the company to disable both models, halting a promotional campaign that had been built around the new releases.
The industry reaction was swift. A coalition of tech and cybersecurity executives issued a public letter demanding that the restrictions be lifted, arguing that the measures would disadvantage U.S. firms in the global AI race. Alex Stamos, chief product officer at Corridor, warned that policymakers were overestimating the advantage American labs hold over foreign competitors, noting that the performance gap is measured in months, not years.
Critics also highlighted the practical challenges of enforcing a blanket ban on foreign‑national usage. Legion Intelligence’s co‑founder Ben Van Roo called the directive “impossible to enforce,” pointing out that the AI community’s global nature makes such restrictions largely symbolic.
Anthropic’s predicament arrives at a volatile moment for the U.S. AI sector. OpenAI, Google, and Microsoft have all launched comparable models, and any precedent set by the Trump administration could ripple across the industry. While the administration has expressed a desire to reduce regulatory burdens on tech firms, officials acknowledged that a uniform export‑control policy might eventually target all advanced AI models, not just Anthropic’s.
As of Monday, talks between Anthropic and the government remain unresolved. The company continues to negotiate not only over the export‑control order but also over ongoing disputes with the Department of Defense regarding permissible military uses of its technology. The outcome will likely shape how the United States balances national‑security concerns with the rapid commercialization of powerful AI systems.
This article was written with the assistance of AI.
News Factory APP - agentic news to boost your SEO & AEO.