U.S. lifts export controls on Anthropic AI models, ending feud

The Commerce Department lifted export controls on Anthropic's Claude Fable 5 and Mythos 5 on June 30, ending a standoff that had forced the company to disable both models worldwide. Anthropic said it would start restoring access Wednesday, bringing Fable 5 back to global users on Claude.AI, Claude Code and the broader Claude platform, with access on Amazon Web Services, Google Cloud and Microsoft Foundry to follow as soon as possible. Mythos 5 had already been partially restored for some U.S. organizations after approval on June 26.

The reversal marked a sharp shift from the June 12 order that required Anthropic to suspend access for any foreign national, including foreign-national employees, which pushed the company to shut the models down everywhere. Fable 5 is the consumer version of Mythos 5, with additional safeguards meant to reduce misuse, and Anthropic had positioned both systems among its most capable frontier models before the government intervened.

Commerce Secretary Howard Lutnick said Anthropic no longer needed an export license after agreeing to proactively detect and address security risks, work with the government on protocols for future releases and report malicious activity it finds in its models. Anthropic and the administration have also discussed a technical assessment to standardize how similar model risks are measured in the future. The company said it will continue working with the government through its Glasswing cybersecurity program, leaving Washington with a more targeted set of guardrails even as the models return to wider use.

The fight became a test case for how Donald Trump’s administration is trying to balance AI cyber risks against competition with China. Tech executives and investors warned that pulling Anthropic’s most advanced models from public use gave Chinese open-source rivals time to catch up, while a June 15 open letter signed by 76 cybersecurity experts, including Alex Stamos, Casey Ellis, Jon Callas, Paul Vixie, Dino Dai Zovi, Katie Moussouris and Rachel Tobac, pressed the White House to lift the controls so defenders could keep finding weaknesses before attackers did. OpenAI also agreed the week before to limit one of its advanced models under White House pressure, showing the Anthropic dispute was part of a broader policy reset, not an isolated clash.

The rollback restores access to some of the most closely watched U.S. frontier AI systems after nearly three weeks of disruption, but it also leaves Commerce with a clearer playbook for demanding security commitments when powerful models cross borders.