The White House has ordered AI firm Anthropic to block all foreign nationals from its advanced Mythos 5 and Fable 5 models, citing national security risks after a "jailbreak" method was reportedly discovered. Anthropic complied by disabling access for all users. The move, a sharp turn in the Trump administration's AI policy, has sparked alarm among U.S. allies who rely on the technology for their own cybersecurity.
Anthropic export control directive
- ▪The government action was prompted by concerns that another company had found a method to bypass, or "jailbreak," the safeguards of the Fable 5 and Mythos models
- ▪To comply with the directive, Anthropic disabled access to the Fable 5 and Mythos 5 models for all of its customers
- ▪According to the Wall Street Journal, concerns raised by Amazon CEO Andy Jassy to the White House instigated the government's directive to Anthropic
- ▪The directive applies to any foreign national, whether inside or outside the United States, including Anthropic's foreign national employees
- ▪The U.S. government directed AI company Anthropic to block all non-Americans from using its Fable 5 and Mythos 5 models
Trump administration AI policy shifts
- ▪Earlier in his second term, President Donald Trump issued an executive order that rescinded a 2023 Biden-era policy on "Safe, Secure, and Trustworthy" AI
- ▪On June 2, 2026, the Trump administration published an executive order to create a framework for “voluntary” federal oversight of AI
- ▪The Trump administration's action represents a significant shift from its earlier, more "laissez-faire" approach to AI regulation
- ▪President Trump had previously ordered government agencies to stop using Anthropic's models after the company rejected Defense Department demands for unrestricted access
Allied government reliance concerns
- ▪Many close U.S. allies were already using Anthropic's Mythos model to identify cybersecurity vulnerabilities in their critical systems
- ▪The U.S. government's action has raised alarms in foreign capitals about the reliability of American AI for critical applications
Amodei's role in escalation
- ▪On June 10, 2026, two days before the directive, Anthropic CEO Dario Amodei called for mandatory third-party testing of AI model risks
- ▪Some critics, including former Meta AI head Yann Lecun, blamed the government's action on Amodei’s “ridiculous fear mongering” about AI threats
Story comments
Loading comments…