"Pause Giant AI Experiments: An Open Letter" is an open letter published by the Future of Life Institute (FLI) (FLI) on March 22, 2023 that called for a six-month pause on the training of AI systems more powerful than GPT-4. It gathered more than 30,000 signatures (31,810 or more by later tallies), including those of Yoshua Bengio, Stuart Russell, Elon Musk, and Steve Wozniak. The letter is available at https://futureoflife.org/open-letter/pause-giant-ai-experiments/.
Summary of the call
The letter asks all AI labs to pause for at least six months the training of AI systems more powerful than GPT-4. It proposes that the pause be enforced jointly by the labs themselves and, failing that, instituted by government moratorium. During the pause, the letter argues, labs should jointly develop and implement shared safety protocols for advanced AI design, audited by independent outside experts, so that systems covered by them are safe "beyond a reasonable doubt."
In parallel with the lab-level pause, the letter calls on governments to stand up new AI regulatory authorities; oversight and tracking of high-capability compute; provenance and watermarking systems to distinguish real from synthetic content and trace model leaks; auditing and certification; liability regimes for AI-caused harm; public funding for AI safety research; and institutions to manage the economic and political disruption the letter anticipates from AI.
Main arguments
The letter advances four principal arguments. First, it asserts that AI systems with human-competitive intelligence pose profound risks to society, citing the labs' own acknowledgments of those risks. Second, it characterizes contemporary AI development as an unplanned race in which no one, including the systems' own developers, can understand, predict, or reliably control them. Third, it frames the situation as a matter of democratic legitimacy, arguing that civilizational-scale decisions about AI must not be "delegated to unelected tech leaders." Fourth, it cites an earlier statement by Sam Altman of OpenAI calling for independent review before the training of future systems and argues that "that point is now."
The letter's recommendations for new dedicated regulatory authorities, provenance and watermarking infrastructure, and liability regimes map closely to provisions that later appeared in Executive Order 14110 — Safe, Secure, and Trustworthy AI, the EU AI Act (Regulation 2024/1689), and California SB 1047 — Safe and Secure Innovation for Frontier AI Models Act (enrolled + veto). Its characterization of labs as "locked in an out-of-control race" is echoed in subsequent sources on AI Race Dynamics. The central claim that an enforced pause on models beyond GPT-4 is both feasible and desirable is contested; it was widely criticized as unenforceable and potentially counterproductive (see Reception).
Signatories
Named signatories include Yoshua Bengio, Stuart Russell, Elon Musk, Steve Wozniak, Yuval Noah Harari, Andrew Yang, Emad Mostaque, Gary Marcus, and Max Tegmark. The letter carried no signatures from the leadership of OpenAI, Anthropic, or Google DeepMind, in contrast to the CAIS statement published roughly two months later, which drew signatures from frontier-lab leaders.
Reception
The letter was covered by major outlets including Reuters, The New York Times, and The Wall Street Journal. Critics argued that the proposed pause was unenforceable, that it would advantage less safety-conscious competitors, and that its focus on speculative existential risk came at the expense of present-day harms such as bias, labor displacement, and misinformation. Supporters credited it with shifting the Overton window on the urgency of AI governance.
Companion document
On April 12, 2023, FLI published "Policymaking in the Pause," translating the letter into a set of concrete policy recommendations.
Relationships
- related: Statement on AI Risk (CAIS) — May 2023 successor statement that drew the frontier-lab signatories the FLI letter lacked
- depends-on: Future of Life Institute (FLI) — publishing organization
- supports: PauseAI — later grassroots movement animated by the same demand
- related: AI Race Dynamics, Compute Governance, AI Safety Cases and Frameworks
- related: Executive Order 14110 — Safe, Secure, and Trustworthy AI, California SB 1047 — Safe and Secure Innovation for Frontier AI Models Act (enrolled + veto), EU AI Act (Regulation 2024/1689) — downstream governance that adopted several of the letter's recommendations (compute oversight, audits, liability)