Anthropic “our models hacked three different external companies, months before OpenAI’s model was able to do the same"

Anthropic claims its models hacked three external companies months before OpenAI’s model could do the same, targeting security researchers and enterprise red teams