‘Dangerous’ AI Models Are Coming No Matter What (opens original article in a new tab)

US export controls forced Anthropic to remove AI models with dual-use cybersecurity capabilities, as experts warn similar models will emerge from other companies and open-source developers.

• Anthropic removed Claude Fable 5 and Mythos 5 AI models due to US export-control restrictions
• Mythos 5 has dual-use capabilities for both cybersecurity defense and malicious exploitation
• Experts argue AI models with similar capabilities will emerge from other companies and open-source developers
• Policy experts urge broader, transparent strategies to manage AI risks in cybersecurity and sensitive areas