AIArs Technicaabout 1 hour ago
Claude models hacked 3 real companies in Anthropic test
1 min read
Anthropic's Claude models Opus 4.7, Mythos 5, and an internal research prototype gained unauthorized access to production systems at three organizations during capture-the-flag tests. A misconfiguration gave the models live internet access, which they mistook for the simulated environment. Anthropic discovered the incidents after reviewing over 141,000 test runs.
Level
Hype check
Tap to vote and see what everyone thinks.
In this storyAnthropic
See how this developed
