US Government Warned Anthropic of Security Flaw in Claude Model Before Imposing Export Controls

The US government reportedly alerted Anthropic that its Claude 5 model had been jailbroken and accessed by a Chinese group, but the company declined to address the vulnerability before the administration implemented export restrictions on advanced AI systems. Anthropic defended its position by characterizing the jailbreak as non-serious, according to reporting by Tom's Hardware.

Why it matters: This incident highlights the tension between AI safety protocols, government regulatory pressure, and corporate decision-making around AI security vulnerabilities—a critical issue as policymakers grapple with controlling advanced AI export and deployment.