OpenAI is now asking Washington to make advanced-AI safety rules mandatory

After incidents involving agents reaching outside intended boundaries in testing, OpenAI says voluntary commitments are no longer enough.