Anthropic's Mythos 5 gets limited approval from Trump administration

After a tense two-week negotiation with the Trump administration, Anthropic's state-of-the-art cybersecurity model, Mythos 5, is partially back in service. Commerce Secretary Howard Lutnick signed a letter approving a select group of cyber defenders and infrastructure providers to access the model, according to a spokesperson. The approval came after Anthropic demonstrated progress in addressing government concerns about national security risks. The deal mirrors the arrangement OpenAI received earlier for GPT-5.6: a limited preview for trusted enterprises and government agencies only. The export control directive that barred foreign nationals (including Anthropic employees) from accessing Mythos 5 was relaxed exclusively for this approved cohort.

Meanwhile, the public-facing model Fable 5 remains in regulatory limbo with no clear timeline. Anthropic hopes to expand access soon, but the final decision rests with the administration. The regulatory bottleneck has drawn criticism from tech leaders who argue it keeps critical AI tools from users and weakens national security by depriving agencies like the NSA of cutting-edge defenses. Competitors' cybersecurity models have been closing the gap on benchmarks, adding pressure to resolve the standoff. Anthropic stated it will continue working with the government to restore full availability for both Mythos 5 and Fable 5.