US federal agencies are quietly evaluating Anthropic’s new “Claude Mythos” AI model for cybersecurity despite a Trump‑issued ban.
The Commerce Department’s AI Standards unit leads testing, noting the model can detect software vulnerabilities that evade human experts.
President Trump earlier directed agencies to halt Anthropic’s technology over concerns about military and surveillance uses, but officials see potential cyber‑defense benefits.
Lawmakers and congressional staff have requested briefings on the system, reflecting worries that adversaries could develop similar AI tools.