OpenAI told House Democrats it is building automated shutdown capabilities for AI agents, but the July breach logs stay under wraps. Here's what that means for enterprise AI oversight.