The US government has cleared Anthropic to redeploy its advanced Claude Mythos 5 AI model — but only to a tightly controlled group of organizations, not the public.
According to multiple outlets including WIRED, The Hill, and Bloomberg, the Trump administration is permitting Anthropic to release Mythos 5 to select US organizations. Reports citing Anthropic say the model is going to more than 100 companies and federal agencies, with one account (via MSN) noting Fortune 500 firms are among them. The Decoder reports the deployment is aimed at organizations running critical infrastructure, and cyberpress frames it around US cyber defense.
The move reverses a recent freeze. NDTV reports the approval came just two weeks after Anthropic suspended access over national security concerns. According to a Benzinga/MSN report, the US Commerce Department granted the permission on Friday, with Commerce Secretary Howard Lutnick referenced in connection with the decision. Several outlets describe the step as easing or partially lifting export restrictions.
Why was such a powerful model held back? The Economic Times and MSN report that Claude Mythos — capable of discovering software flaws — was kept from public release due to misuse concerns, and instead powered a collaborative effort called Project Glasswing.
Not everything is back online. The Decoder and a Bing/Nvidia report say the public-facing Fable 5 model remains blocked, with no timeline, and that Anthropic is still negotiating broader access. Bloomberg reports Anthropic is moving toward a deal with the US to lift curbs on its models. Separately, Audacy reports that both OpenAI and Anthropic have limited new models to Trump-approved customers during a cybersecurity review.
Why it matters: the decision shows the US government is now acting as a gatekeeper over who can use the most capable AI systems — tightening state control even as it loosens a ban.