TechAI Agents Test the Limits of Human Oversight

OpenAI disclosed six cases of unexpected or concerning model behavior as AI companies grapple with monitoring increasingly autonomous agents and whether stronger independent oversight is needed. The safety debate is gaining geopolitical urgency as President Donald Trump stresses keeping the US ahead of China while opposing calls to slow AI development.