A model billed as the most capable cybersecurity AI on the market lasted just three days in public hands. Anthropic launched Claude Fable 5 on 9 June, only for the US government to order its global withdrawal after concerns that its automated hacking capabilities had outpaced existing regulatory safeguards.
A Three-Day Window, Then a Global Shutdown
Claude Fable 5 was positioned as a safety-filtered public release built on the same underlying technology as Claude Mythos 5, a more powerful, unrestricted version reserved for government agencies under Anthropic's Project Glasswing. Three days after launch, US authorities suspended access to both systems.
The restriction applies specifically to foreign nationals operating within the United States, including researchers employed directly by Anthropic. Because export-control rules covered its own staff, the company chose to withdraw customer access to the models worldwide on 12 June in order to stay compliant, rather than restrict access selectively.
Security Operations at Machine Speed
James Hadley, Founder and Chief Innovation Officer at Immersive, says organisations are under growing pressure to adopt AI across the workforce while simultaneously turning their internal security operations into agentic systems.
He notes that AI paired with an experienced security researcher can chain suspected vulnerabilities into proven exploits at a scale individual researchers have never matched. For regulated industries in particular, he argues, businesses now need to prove that staff can use AI securely and that security agents can operate at machine speed without runaway token costs.
"AI, especially when coupled with an experienced security researcher, is able to work at a scale and volume not seen before"— JAMES HADLEY, FOUNDER AND CHIEF INNOVATION OFFICER, IMMERSIVE
White House Calls for Remediation
The directive requires Anthropic to block foreign nationals from accessing both Fable 5 and Mythos 5. According to the company, officials believe they uncovered a way to "jailbreak", or circumvent the built-in restrictions of, Fable 5. Anthropic reviewed a demonstration of the bypass and says it exposed only a small number of minor flaws it already knew about, ones it claims rival publicly available tools can also surface without specialist exploits.
David Sacks, White House Adviser and Co-Chair of the Council of Advisors on Science and Technology, said officials issued the export control reluctantly, and claimed Dario Amodei, Anthropic's CEO, declined to fix the jailbreak or pull the model voluntarily. Sacks added that the administration hopes Anthropic will resolve the safety issue so the export control can be lifted and Fable can return to general release.
Notably, Anthropic had said before launch that the technology was too powerful to release. Some observers dismissed that framing at the time as investor-courting marketing ahead of the company's anticipated IPO — a claim the brief, three-day availability window may now lend some credibility to.