Anthropic Suspends Claude Fable 5 and Mythos 5 Over US Government Security Order

Summary

Anthropic has suspended access to its newly released Claude Fable 5 and Mythos 5 AI models following a US government security order, marking a significant escalation in regulatory pressure on the AI industry. The company stated that authorities ordered it to restrict foreign nationals from using the models due to concerns about jailbreaking vulnerabilities—techniques that could allow unauthorized access to sensitive systems. While US officials did not publicly identify specific security threats, Anthropic acknowledged that a method of bypassing the model's restrictions had been demonstrated, though the company characterized the vulnerabilities as minor and consistent with those found in other public models.

The suspension reflects growing tension between Anthropic and the Trump administration, which is already engaged in a separate lawsuit against the company over restrictions on government agency access to its tools. The decision has prompted international concern, with the European Union citing the incident as evidence of Europe's need for technological sovereignty and independence from American AI infrastructure. Security experts, including researchers from the UK's AI Security Institute, noted that Claude Fable 5 demonstrated a significant leap in AI capabilities—the model could exploit system defenses 73% of the time during testing—underscoring the dual nature of advanced AI as both a powerful tool and a potential security risk.

• Claude Fable 5 represents a significant capability leap in AI systems, able to exploit cyber defenses at a 73% success rate, raising questions about whether such models should be widely accessible