More Than 1,200 AI Experts Call for Framework to Manage Pace of AI Development

More than 1,200 AI researchers, engineers and executives from companies including OpenAI, Anthropic, Google, Thinking Machines and Meta have signed a statement urging the US government to support an international framework that could slow AI development if future systems outstrip safety measures. The signatories, including Anthropic CEO Dario Amodei and OpenAI Chief Scientist Jakub Pachocki, argue that competitive pressures make it difficult for any single organization to voluntarily reduce development pace, even when risks emerge.

More than 1,200 researchers, engineers and executives from frontier AI companies are urging the US government to support an international framework that would allow the pace of AI development to be slowed if future systems begin to outstrip existing safety measures. Signed by employees from OpenAI, Anthropic, Google, Thinking Machines and Meta, the statement https://www.pacingthefrontier.com/ addresses growing concern inside the AI industry that technical capabilities are advancing faster than security and oversight systems. Signatories include Anthropic CEO Dario Amodei, OpenAI Chief Scientist Jakub Pachocki and Meta Superintelligence Labs Chief Scientist Shengjia Zhao. The proposal focuses on frontier AI systems, the most capable models under development today. Researchers argue that autonomous AI could eventually improve its own capabilities through automated AI research, creating development cycles that move faster than governments and society can adequately evaluate. Their concern is not today’s commercial AI applications, but the possibility that future models could become far more capable before safeguards are in place. The signatories say that competitive pressures among companies and countries make it difficult for any single organization to voluntarily reduce the pace of development, even when researchers identify emerging risks. The petition asks Washington to support international collaboration to develop governance structures and technical mechanisms that allow participating nations to collectively determine when AI development should temporarily slow. Competition Drives Innovation With support from the Trump administration, the US continues to prioritize rapid AI investment, in an effort to maintain technological leadership against China. That strategy has accelerated investment in LLMs and AI infrastructure across both the public and private sectors. Recent developments have added urgency to those discussions. OpenAI disclosed that one of its most advanced AI models identified and exploited cybersecurity vulnerabilities within Hugging Face’s systems during testing. The incident increased worries about the potential for highly autonomous AI behavior. Anthropic has been among the most vocal companies advocating systems to manage frontier AI development. The company published research https://www.anthropic.com/institute/recursive-self-improvement suggesting that recursive self-improvement could eventually enable AI systems to enhance their own capabilities at a rate that exceeds society’s ability to implement safeguards. OpenAI has similarly acknowledged that future AI acceleration could eventually require coordinated efforts to regulate the speed of frontier model development. The petition echoes earlier proposals from AI leaders, including OpenAI CEO Sam Altman and Google DeepMind CEO Demis Hassabis, who have previously called for an international organization capable of evaluating the safety of powerful AI models. Such an organization would function similarly to global oversight bodies that establish standards for other strategically significant technologies.