The U.S. government moved to restrict Anthropic’s access and distribution of its advanced AI models after a security vulnerability review escalated into export-control action. Reporting tied the decision to a warning from Amazon CEO Andy Jassy and concerns about unauthorized access, including potential cybersecurity risks tied to the Anthropic Fable 5 and Mythos 5 models. According to the account, Amazon researchers used prompts to elicit restricted cyberattack information, then administration officials discussed the issue with Anthropic. Anthropic argued the jailbreak involved a narrow bypass and said it had been given limited time—about 90 minutes—to pull the newest model before Commerce Department export controls barred distribution to foreign nationals, including non-U.S. citizens working inside the country. The episode underscores how higher education and research organizations that rely on frontier AI systems may face sudden availability changes based on national security policy. Campuses and labs building AI workflows for research, privacy-sensitive projects, and education delivery will need rapid contingency plans when model access is curtailed through compliance directives.