U.S. Export Controls Force Anthropic to Offline Claude Fable 5 and Mythos 5

Summary

Anthropic has taken its Claude Fable 5 and Mythos 5 AI models offline following a U.S. government export-control directive barring foreign nationals from accessing the services. The models possess advanced dual-use capabilities for identifying and exploiting software vulnerabilities—capabilities that could benefit cybersecurity professionals but pose risks in the hands of malicious actors. The Trump administration moved to restrict both models, believing Fable 5's safeguards can be circumvented to unlock Mythos 5's full capabilities, citing national security concerns.

However, security experts and researchers argue that the government's regulatory approach misses a fundamental reality: multiple AI companies and open-source developers are already developing or possess comparable capabilities. Experts like Tarah Wheeler (Chief Security Officer at TPO Group) and Bruce Schneier (Harvard/University of Toronto researcher) contend that Anthropic is merely the visible target of a broader technological trend. They warn that existing models, combined with sophisticated prompting techniques, can match or replicate Mythos/Fable's capabilities within months—and that competing companies are likely holding similar tools in reserve while observing how regulators treat Anthropic.

• Security experts argue that similar or matching capabilities are inevitable across the industry and likely already exist at competing companies
• The broader issue is not about individual models but the inevitable proliferation of advanced AI capabilities within 6–24 months