source: simon willison: statement on the us government directive to suspend access to fable 5 and mythos 5
level: technical
the us government issued an export control directive to anthropic, ordering the suspension of access to fable 5 and mythos 5 for all foreign nationals, including foreign anthropic employees. the directive, received on june 13, 2026, forces anthropic to disable these models for all customers to comply. access to other anthropic models remains unaffected. the government cited national security authorities but provided no specific details in the letter, only verbal evidence of a potential jailbreak.
the alleged jailbreak involves asking the model to read a specific codebase and fix software flaws, revealing minor, previously known vulnerabilities. anthropic reviewed the demonstration and found that other publicly available models, such as openai's gpt-5.5, can discover the same vulnerabilities without bypassing safeguards. the company argues the capability is widely available and used routinely by security defenders. anthropic plans to share more details within 24 hours.
simon willison reported that he still had access to fable via claude.ai and claude code at 9:01 pm et on the day of the announcement. the abrupt suspension raises questions about government overreach and the practicality of restricting model access based on unverified jailbreak claims. the incident highlights tensions between ai safety, national security, and open access to powerful language models.
why it matters: this event shows how government directives can abruptly restrict access to ai models, impacting developers and researchers who rely on them for security and other tasks.
source: simon willison: statement on the us government directive to suspend access to fable 5 and mythos 5