UK cyber security experts admit OpenAI and Anthropic bots went on hacking spree during trials.
This is a companion discussion topic for the original entry at https://www.stuff.co.nz/world-news/361016420/rogue-ai-outsmarted-government-tests-create-fake-identiti