Anthropic Pulls AI Models Offline Following US Export Control Order

Anthropic took its Claude Fable 5 and Mythos 5 AI models offline after a US government export-control directive barred foreign nationals from using them, and the company is in talks with the White House. The models possess dual-use capabilities, able to both identify and exploit software vulnerabilities, posing potential risks if accessed by malicious actors. Initially, Mythos 5 was privately released to a select working group, while the public version of Claude Fable 5 had restrictions on biology and cybersecurity topics.