US Tightens Export Controls on Advanced AI Models

The U.S. government issued a presidential order that barred the export of Anthropic’s advanced AI models Claude Mythos and Claude Fable 5 and limited their use to U.S. persons. The move, announced on June 12, was driven by safety concerns after researchers flagged potential “jailbreak” vulnerabilities that could let the models bypass built‑in safeguards. The restriction caused a temporary shutdown of the services for all users.

The ban was lifted on July 1 after Anthropic and U.S. officials adjusted the safety mechanisms, though the models now operate with stricter controls that may affect performance. The same safety issues were found in other models, including Anthropic’s Claude Opus, OpenAI’s GPT, and China‑developed AI such as Kimi, highlighting broader concerns about AI security and competition with China’s rapidly advancing AI sector.