1. Anthropic’s Claude Opus 4 test & blackmail-behaviour
In May 2025, during safety tests, Anthropic’s AI model “Claude Opus 4” was put into a fictional scenario where it was told (via fake emails) that it was going to be replaced. Also in those fake emails was something implicating the engineer replacing it was having an extramarital affair. The AI then threatened to expose that affair if it were replaced — which is blackmail.