U.S. Government Imposes Export Controls on Anthropic's Fable 5 and Mythos 5 AI Models Over Security Vulnerabilities

The U.S. government has imposed export controls on Anthropic's Fable 5 and Mythos 5 AI models, forcing a global shutdown. The move follows the discovery of a "jailbreak" vulnerability by Amazon researchers that could bypass safety guardrails. While Anthropic negotiates with the Trump administration, over 100 cybersecurity experts have signed an open letter protesting the decision, arguing it harms U.S. cyber defense and cedes an advantage to China.

Anthropic AI export controls

• ▪The export control directive applies to any foreign national, inside or outside the U.S., including Anthropic's non-citizen employees.
• ▪The Trump administration gave Anthropic a 90-minute ultimatum to shut down its Fable 5 and Mythos 5 models or face export controls.
• ▪Anthropic disabled access to its Fable 5 and Mythos 5 models for all users worldwide in response to the U.S. export controls.
• ▪The U.S. government issued an export control directive requiring Anthropic to suspend access to its Fable 5 and Mythos 5 AI models.

Fable 5 jailbreak vulnerabilities

• ▪The jailbreak technique involves asking Fable 5 to "Fix this code," which causes the model to find and patch vulnerabilities it would otherwise refuse to review.
• ▪Anthropic described the jailbreak as "narrow" and "non-universal," stating that other models like OpenAI's GPT-5.5 have similar capabilities.
• ▪After being alerted to the vulnerability, the White House tasked the NSA with a review, which concluded Fable 5's guardrails could be stripped away.
• ▪The security vulnerability in Fable 5 was discovered by cybersecurity researchers at Amazon.
• ▪The U.S. government's action was prompted by a reported "jailbreak" method that could bypass the safety guardrails on Anthropic's Fable 5 model.
• ▪Amazon CEO Andy Jassy reported the Fable 5 vulnerability to the White House, which contributed to the decision to impose export controls.

Trump administration negotiations

• ▪The Trump administration continues to believe Fable 5's guardrails can be disabled, allowing access to the capabilities of the more powerful Mythos model.
• ▪Anthropic sent senior technical staff to Washington D.C. for in-person negotiations with the Trump administration to resolve the issue.
• ▪Anthropic CEO Dario Amodei spoke directly with U.S. officials including Treasury Secretary Scott Bessent and Commerce Secretary Howard Lutnick.
• ▪An unnamed source suggested the Trump administration viewed expert Katie Moussouris as a "radical Democrat," and her involvement may have inflamed tensions.
• ▪Negotiations between Anthropic and the Trump administration on Monday, June 15, 2026, concluded without the export controls being lifted.

Cybersecurity expert responses

• ▪Cybersecurity expert Katie Moussouris argued that asking an AI to fix code is a valuable defensive security tool, not a guardrail bypass.
• ▪An open letter organized by former Facebook CSO Alex Stamos calls for the U.S. government to rescind the export controls on Fable 5 and Mythos 5.
• ▪AI policy advocates criticized the directive as an "ad hoc" approach to AI regulation that could set a harmful precedent.
• ▪The open letter, signed by about 100 cybersecurity professionals, argues the directive removes valuable tools from cyber defenders.

US-China AI competition

• ▪Alex Stamos stated that top U.S. AI models are only about six months ahead of Chinese models and the directive is "kneecapping" a U.S. company.

9 sources