US Export Controls Force Anthropic to Take Down Advanced AI Models

Anthropic took its Claude Fable 5 and Mythos 5 AI models offline after a US government directive barred foreign nationals from using them, and the company is still negotiating with the White House to restore access. The models are considered dangerous due to their dual-use capabilities in cybersecurity, able to both find and exploit software vulnerabilities. Anthropic initially limited access to Mythos Preview through a select working group called Project Glasswing, and while Mythos 5 was privately released, Claude Fable 5 was made public with restrictions on biology and cybersecurity queries.