Following a recent restriction, the US Commerce Department has approved the use of Anthropic’s Mythos 5 model by more than 100 government agencies and companies, including non-American employees at those organizations, marking a softened regulatory stance on sensitive AI cybersecurity tools.
- Mythos 5 access restored for 100+ US agencies and companies
- Non-American employees of approved organizations now included
- Fable 5 remains restricted pending further review
What happened
Two weeks after the Trump administration imposed a ban on Anthropic's cybersecurity-driven AI models Mythos 5 and Fable 5 due to concerns about easily bypassed safety guardrails, the government has announced a partial easing of those restrictions. Over 100 specific US government agencies and companies are now authorized to use Mythos 5.
Why it matters
The move signals a cautious but significant shift in regulatory approach toward advanced AI tools with cybersecurity applications. By enabling critical US infrastructure operators and trusted partners to leverage Anthropic’s strongest security model, the government recognizes the potential benefits of integrating cutting-edge AI while managing associated risks.
Including non-American employees in the access list also highlights a pragmatic stance on operational realities within multinational organizations and government contractors. Nonetheless, ongoing concerns about AI model vulnerabilities keep Fable 5 under continued scrutiny, reflecting the complex balance between innovation and controlled deployment.
What to watch next
Industry observers and stakeholders should monitor how broadly Mythos 5 adoption expands within the approved organizations and whether this oversight framework successfully mitigates security risks that prompted the initial ban. Anthropic’s stated intention to continue working with the government to further reopen Fable 5 signals potential future shifts in access policy.
Additionally, the effectiveness of these safeguards and official assessments may influence regulatory approaches for similar AI models in the cybersecurity sector, shaping how advanced AI is integrated into critical national infrastructure operations going forward.