Read, Watch or Listen

Washington — The U.S. government on Friday partially lifted its export block on Anthropic’s most advanced artificial intelligence system, Claude Mythos 5, ending a two‑week standoff rooted in national security concerns that began on June 12, 2026. In a letter to Anthropic’s chief compute officer, Tom Brown, Commerce Secretary Howard Lutnick said “appropriate safeguards” are now in place to let a group of about 100 trusted U.S. companies, federal agencies, and institutions regain access to the model. Anthropic said it is quickly restoring Mythos 5 for organizations that operate and defend critical infrastructure, describing the system as its strongest cybersecurity model. Many of the approved entities are reportedly members of Project Glasswing, Anthropic’s coalition of prominent technology companies and institutions, although the Commerce Department has not released the full list of authorized users. Washington — The partial reversal follows Anthropic’s abrupt June 12 shutdown of both Mythos 5 and its public counterpart, Fable 5, after the government ordered the company to suspend access for all foreign nationals, including non‑U.S. employees, under “national security authorities.” Fable 5 remains offline for the public, and Lutnick’s letter did not permit its redeployment, as officials continue negotiations over concerns that users could jailbreak the system to identify software vulnerabilities. The move comes the same day rival OpenAI announced three new models and agreed to limit rollout of GPT‑5.6 at Washington’s request, a parallel step that underscores what observers see as the emergence of a formal U.S. regulatory regime for frontier AI. Civil liberties advocates have criticized the opaque selection of beneficiaries, while U.S. allies and foreign organizations remain barred from Mythos 5 under the current licensing framework, which only allows specific exceptions named in the Commerce Department’s letter.

Prepared by Christopher Adams and reviewed by editorial team.