The Trump administration moved to restrict access to Anthropic’s newest AI systems after security concerns escalated between industry and federal officials. Multiple reports say Amazon CEO Andy Jassy raised concerns following prompts used by Amazon researchers to probe cybersecurity information tied to Anthropic’s Mythos-class model. By Friday evening, the Commerce Department used national security export controls to bar Anthropic from distributing Fable 5 and its underlying Mythos 5 to foreign nationals, including certain non-citizens inside the U.S. Anthropic said its safeguards were bypassed in a narrow way, arguing the incident did not amount to a full jailbreak. Canadian Prime Minister Mark Carney warned that such moves underscore the risk of overreliance on a small number of American providers, while Luta Security CEO Katie Moussouris disputed the government framing, characterizing the research as “Defense Oriented Prompting.” The episode highlights how export-control policy is increasingly shaping who can access frontier AI capabilities and under what conditions.