20 AI Researchers Warn of Extreme Risks from Automated AI Research

On September 28, 2026, 20 prominent AI researchers, including Geoffrey Hinton and Yoshua Bengio, issued a warning regarding the extreme risks associated with the automation of AI research and development. They emphasized that advancements could shift from years to mere months, significantly accelerating the pace of AI capabilities. Jakub Pachocki, a research lead at OpenAI, stated that no laboratory has achieved sufficient alignment to responsibly scale AI research at maximum speed for an extended period. Additionally, 42 mathematicians have joined the call for increased attention to these risks, highlighting a growing consensus in the academic community. Reports suggest that some employees at Anthropic are already exploring safe havens in anticipation of potential adverse outcomes from AI developments. This warning underscores the urgency for practitioners to prioritize safety and alignment in their ongoing AI projects.