Turns out it really was Amazon who snitched. Last Thursday, Amazon's CEO reported to the U.S. government that Claude's Fable model has the capability to identify software vulnerabilities. Then, the very next morning on Friday, the U.S. Treasury Secretary and other White House officials held a meeting to discuss the matter. Shortly after, the White House started reaching out to Anthropic's CEO, engaging in three rounds of tense discussions. By Friday night, the U.S. government urgently issued export controls, restricting foreign access to Anthropic's most advanced Claude model. The reason? This thing can actually auto-read software codebases, identify security vulnerabilities, and even suggest attack strategies. In the end, it was urgently restricted.