The world's most powerful AI models experienced a significant disruption due to a dispute with the Trump administration. This occurred when Amazon, a partner and investor in Anthropic, reported a "jailbreaking" vulnerability in Anthropic's latest models, Mythos and Fable. Amazon's concerns triggered sweeping U.S. export controls, leading to the models being taken offline for twenty days. Cybersecurity experts later argued that similar issues existed in other leading AI models.Commerce Secretary Howard Lutnick informed Anthropic CEO Dario Amodei about the impending export controls, clarifying that the intention was to get the models removed from public access. This prompted a three-week intensive effort involving multiple government agencies and Anthropic engineers to address the perceived safety concerns. Despite Anthropic's efforts and technical refinements, federal agencies initially deemed the changes insufficient.The situation intensified due to personality clashes and communication breakdowns within the government and with Anthropic. Treasury Secretary Scott Bessent played an early role in raising the alarm and facilitating re-engagement with Anthropic. Various government officials, including those from the National Security Agency and the White House Office of Science and Technology Policy, participated in the technical discussions.Ultimately, after further modifications and approvals, Anthropic's models were re-released on July 1. However, the event highlighted the complexities of AI safety regulation and raised questions about the future release of these models to allied countries. The incident also underscored the lack of a clear, transparent framework for approving future AI models across the industry.