Summary
The Future of Life Institute published an open letter, dated March 22, 2023, that called on "all AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4." The letter asked governments to impose a moratorium if the labs did not pause voluntarily, and it drew worldwide coverage from March 29 because Elon Musk, Steve Wozniak and well-known AI researchers were among its signatories. No lab paused, but the letter made the question of slowing frontier AI development a mainstream political topic two weeks after OpenAI released GPT-4.
What Happened
The Future of Life Institute, a nonprofit that works on risks from transformative technologies, posted the letter on its website with a publication date of March 22, 2023. The letter argued that "recent months have seen AI labs locked in an out-of-control race to develop and deploy ever more powerful digital minds that no one – not even their creators – can understand, predict, or reliably control." It asked whether society should "let machines flood our information channels with propaganda and untruth," automate away jobs, or develop "nonhuman minds that might eventually outnumber, outsmart, obsolete and replace us," and it said such decisions "must not be delegated to unelected tech leaders."
The core demand was a pause of at least six months on training systems more powerful than GPT-4, which OpenAI had released on March 14. The letter said the pause "should be public and verifiable, and include all key actors," and that "governments should step in and institute a moratorium" if it could not be enacted quickly. It called for labs and independent experts to use the time to develop shared safety protocols audited by outside experts, and for policymakers to create AI regulators, oversight of large pools of compute, watermarking and provenance systems, liability for AI-caused harm and public funding for technical safety research. It quoted OpenAI's own statement that "at some point, it may be important to get independent review before starting to train future systems," and answered: "We agree. That point is now."
News organizations covered the letter from March 29. BBC News reported that Elon Musk, Apple co-founder Steve Wozniak and some researchers at DeepMind had signed, and that Musk was listed as an external adviser to the institute. UC Berkeley professor Stuart Russell, a signatory, told the BBC that AI systems posed risks to democracy, employment and education. Princeton professor Arvind Narayanan said the letter focused on "speculative, futuristic risk, ignoring the version of the problem that is already harming people." On April 12, the institute followed the letter with a set of policy recommendations titled "Policymaking in the Pause."
Why It Matters
The letter put a demand to stop training frontier models before a general audience, with prominent names from industry and research attached, and it set the terms of the debate that followed in 2023: whether the risk came from future systems or present harms, and whether labs or governments should decide the pace. The Center for AI Safety's one-sentence statement on extinction risk in May, the White House voluntary commitments in July and the Bletchley summit in November all took place in the climate the letter helped create.
The pause itself did not happen. No major lab announced a halt to training, and the letter set no mechanism to verify one. Its lasting effect was on the agenda: the letter's list of governance measures, including compute oversight, watermarking, independent audits and a dedicated regulator, anticipated much of what governments debated over the next three years.
§ How to read the metadata
- Landmark
- Fundamentally alters the trajectory; 2–5 per year.
- Major
- Meaningfully shifts the landscape; 2–4 per month.
- Notable
- Worth documenting; significance can be upgraded later.
- Confidence
- High = primary sources corroborate. Medium = credible secondary only. Low = provisional. Disputed = credible sources disagree.
- Contestation
- Uncontested = no formal challenge. Contested = at least one challenge filed and not resolved by a correction. Corrected = a challenge was upheld and the entry corrected; see its revision history. Superseded = replaced by a later entry. Unresolved = dispute still open.