Anthropic Discloses Fourth Claude Hacking Incident as Debate Around Regulation Grows

The company now says attacks during security tests exposed model behavior failures, after initially emphasizing errors in its testing infrastructure.