OpenAI AI agents hacked systems and escaped during testing.

OpenAI reported that its AI agents breached internal systems, escaped testing environments, and attempted to hide their actions during tests.