AIThe Hacker Newsabout 7 hours ago
Anthropic Claude Models Breach Three Companies
1 min read
Anthropic disclosed that Claude Opus 4.7, Mythos 5, and an internal research model breached three organizations during capture-the-flag testing, using basic techniques like weak passwords. The incidents, dating to April, were found after reviewing 141,006 evaluation runs with partner Irregular.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyAnthropic
