Anthropic took its Claude Fable 5 and Mythos 5 AI models offline after a US export-control directive barred foreign nationals from using them. The models, which can both find and exploit software vulnerabilities, were partially released to a private consortium and the public with restrictions on sensitive topics.
Anthropic acknowledges the dual-use nature, warning that the same capabilities could be dangerous in malicious hands. The company is still negotiating with the White House to potentially reinstate the services.
Comments