Anthropic won U.S. clearance to restore some access to its Mythos 5 cybersecurity model, after the Commerce Department flagged national security risks and required changes to “Covered Models.” Commerce Secretary Howard Lutnick said the company made “significant progress,” allowing redeployment to a small group of “trusted partners,” though access to related Fable 5 remains restricted. The clearance follows a broader attempt to rein in frontier AI capabilities that can be used for cyber operations. Anthropic previously objected to export controls that pushed it to disable global access for foreign users, and it said it is working with the government on an expanded policy framework for future guardrail circumvention. Separately, Anthropic’s latest public dispute alleges Alibaba extracted capabilities of Claude through a cheaper “distillation” approach using fake accounts and low-cost interactions. Leading IPO expert Jay Ritter told Fortune the episode will influence whether investors treat Anthropic’s frontier AI moat as defensible—or as a profitability risk if competition catches up.