The US government has granted Anthropic permission to share a version of its AI model, Mythos, with select government agencies, reversing an earlier decision to restrict access. Initially, the government prohibited Anthropic from making its Fable AI program available outside the US or to non-American users within the country, fearing potential misuse related to identifying software vulnerabilities. Anthropic had globally restricted access to Fable due to difficulties verifying customer nationalities. Concerns centered on the possibility of identified security flaws being exploited against the government itself. Following discussions with the government regarding built-in safety protocols, a compromise was reached allowing limited access to Mythos for security assessments. The government is also individually vetting access requests for OpenAI’s GPT-5.6, another AI tool specializing in vulnerability detection, demonstrating a broader pattern of caution regarding advanced AI technologies. Anthropic states it will continue collaborating with the government to broaden access to Fable in the future.