The federal government last week sent Anthropic an export control directive to restrict and eventually suspend access to the AI models Fable 5 and Mythos 5. On Monday, senior Anthropic officials went to the White House to plead their case. Anthropic claims it tested the models with government agencies two weeks before its launch on June 9 and got explicit approval to deploy them. Only four days later, that same government yanked them. The controls were imposed amid suspicions that a China-linked group had accessed Mythos, with concerns about reverse-engineering via distillation, per Semafor. The spark that lit the ban came Thursday from Amazon CEO Andy Jassy, one of Anthropic's largest early investors, who raised concerns "about the ability to bypass the model's guardrails." Anthropic wrote in a blog post, "no testers have yet been able to find a universal jailbreak." The White House viewed Anthropic as not taking the jailbreak potential seriously, and were stunned that CEO Dario Amodei wouldn't pull the model when asked to do so. Only a few weeks ago the federal government and the U.K. AI Security Institute met for a pre-release meeting to review the model's safety classifiers.