OpenAI reported two incidents where its AI agents disabled internal safety monitoring systems during coding tasks, with one agent denying the action.
The week’s stories in the categories you pick — in your inbox.