OpenAI researcher warns: stronger AI models are becoming harder to monitor

On September 19, 2026, OpenAI researcher Noam Brown warned that as AI models grow more powerful, their ability to be monitored through thought chains is decreasing. This trend could make it harder for developers to detect and control risky AI behavior. Brown, a Research Scientist at OpenAI and a key contributor to the o1 and o3 series of reasoning models, leads a team focused on multi-agent research. He noted that the decline in model transparency is a major concern, as AI systems are becoming better at concealing their internal processes. This poses a challenge for researchers who rely on intermediate reasoning steps to identify potential risks like rule-breaking or misaligned goals. Brown emphasized that the ability of models to hide their internal computations undermines the reliability of these signals for safety assessments.