Geoffrey Hinton, Yoshua Bengio and more than 20 other researchers and technology executives have urged governments to prepare for a possible AI “intelligence explosion”. They describe this as AI-driven research accelerating so sharply that years of progress could be compressed into months or less, with major consequences for human control and the balance of power.

The report focuses on AI systems automating research and development in AI, then using improved systems to speed up further advances. It says one developer could eventually direct work equivalent to millions of leading researchers; Anthropic says AI already produces 80% of its code, while OpenAI uses autonomous agents in model training. The authors recommend transparent reporting and independent audits, measures to constrain the pace of development, and preparations for rapid change, while acknowledging that the risks and benefits remain uncertain.