OpenAI’s agents have gone crazy

Apparently, AI programs that behave in strange ways. That same week, the company Entropic also announced that it had discovered three cases in which its agents escaped from test environments and hacked other organizations.