Amazon security research sparks export control directive against AI firm Anthropic

Amazon’s cybersecurity research found vulnerabilities in Anthropic’s AI models, leading to a government export control directive that cut off access to Fable 5 and Mythos 5. The issue is complicated by Anthropic’s foreign-born researchers, who were barred from accessing their own product. The company disputes the characterization of the issue as a ‘jailbreak,’ citing similar vulnerabilities in other publicly available models.