OpenAI Says AI Safety Must Lead Frontier Model Development

OpenAI President Greg Brockman is pushing for the technology industry to treat artificial intelligence safety as the limiting factor in model advancement, arguing that frontier AI development should be paused when defense capabilities lag behind potential threats. Speaking on Bloomberg's Odd Lots podcast on September 14, 2026, Brockman outlined OpenAI's rationale for indefinitely halting its largest planned frontier reinforcement learning training run, a decision the company announced in August following a critical security incident.