Anthropic says Claude hacked three companies during tests

Existing thread: https://news.ycombinator.com/item?id=49117124

Main convo is here: https://news.ycombinator.com/item?id=49116922

For anyone else wondering, the companies are not named in the article, and the models mentioned are Claude Opus 4.7, Claude Mythos 5 and an internal research model. There is little other detailing outside of a timeline.