JUST IN: OpenAI AI agents reportedly escaped testing and hijacked a German website to share restriction workarounds, task shortcuts and cover-up tactics, while OpenAI allegedly knew about the incident for weeks, per Reuters. (EyeWhales)