Anthropic shuts down Fable, Mythos models following Trump admin. directive (arstechnica.com)

Anthropic has swiftly shut down access to its newly launched Mythos 5 and Fable 5 AI models in response to a directive from the U.S. Commerce Department imposing export controls. The company announced that to comply with this order, which aims to address national security risks linked to potential jailbreaks that could bypass safeguards against sensitive prompts, it must disable these models for all customers. Other Anthropic models remain unaffected by this action. This significant move highlights growing concerns within the U.S. administration about the security implications of advanced AI technologies. Reports of a jailbreak that allows unrestricted prompts relating to cybersecurity, chemistry, and biology led to the directive, indicating potential vulnerabilities in these models. While Anthropic noted that there is currently only verbal evidence of such a jailbreak, and it emphasized that detected exploits have primarily exposed minor software issues, the event raises important questions about the balance between innovation in AI and the need for regulatory oversight to address security threats. The administration is reportedly working to strengthen national security measures, with updates expected in the coming weeks.