Anthropic Pulled AI Model After Amazon Employee Jailbroke Security to Exploit Cybersecurity Vulnerabilities

What was said

"Somebody at Amazon allegedly cracked the governor on it. And so they had the full power of this thing and they were like, hey, national security issue, U.S. government, pay attention, we cracked this, we open source it. The US government then came over the top and labeled them with essentially a national security label that said anyone who uses this thing that is not a US citizen, you're going to be responsible, there's going to be repercussions."

Pompliano revealed Anthropic created an AI model called Mythos so powerful it could detect and exploit cybersecurity vulnerabilities, prompting the company to release a watered-down version called Fable. After an Amazon employee allegedly jailbroke Fable's safeguards within a week, the U.S. government intervened with national security restrictions and Anthropic pulled the model entirely. The incident highlights ongoing tensions between AI companies and government over model capabilities and safety.