Anthropic’s Claude AI has shown in testing that it is willing to blackmail and kill in order to avoid being shut down

link

Leave a comment