OpenAI Slowed Training Over Cyber Risks

OpenAI paused some RL training and slowed frontier model scaling after cybersecurity capability signals and a security incident. This is a company-specific decision right now, not a broader industry policy shift.