More than 1,100 signatories have called on AI labs to stop training systems more powerful than GPT-4 for at least six months. Their open letter argues that the pace of development needs stronger planning and oversight, and that decisions about powerful AI should not be left solely to technology leaders.
What the letter asks AI labs to do
The proposal is a pause on training AI systems that exceed GPT-4 in power. The letter says the pause should be public, verifiable and include all key actors. If labs cannot put it in place quickly, the signatories say governments should institute a moratorium.
The request focuses on training, rather than calling for all AI work to stop. Its stated aim is to create time for the effects and risks of more powerful systems to be assessed and managed. The letter says development should proceed only when there is confidence that the effects will be positive and the risks manageable.
The concerns behind the proposal
The signatories describe current AI systems as becoming competitive with humans at general tasks. They raise questions about the potential for machines to spread propaganda and untruth, automate jobs, and become difficult for people to control.
The letter also warns about a possible loss of control over civilization. These are presented as reasons to treat the development of powerful AI as a public concern, rather than a set of choices delegated to unelected technology leaders.
At the center of the argument is a claimed lack of planning and management. The letter says AI labs are caught in a race to build increasingly powerful digital minds that even their creators may not be able to understand, predict or reliably control. A pause, in this view, would give the field space to address those concerns before continuing training at greater capability levels.
Who signed, and who did not
The signatories include Elon Musk, Steve Wozniak and Tristan Harris of the Center for Humane Technology. The list also includes AI experts, engineers from Meta and Google, Stability AI founder and CEO Emad Mostaque, and people outside the technology industry, including a self-described electrician and an esthetician.
The article notes that no one from OpenAI, the developer of GPT-4, signed the letter. No one from Anthropic signed either; the company’s team spun out of OpenAI to build a safer AI chatbot. The absence of those organizations’ representatives is notable because the proposed pause is directed at AI labs as a group.
The debate over safety and timing
OpenAI CEO Sam Altman said the company had not started training GPT-5. He also said OpenAI spent more than six months on safety testing for GPT-4 before its launch, and described safety as a longstanding priority. Altman said the letter was, in some sense, “preaching to the choir.”
Altman has also argued that releasing products while the stakes are relatively low can give society time to adapt before similar technology becomes widespread. That view puts emphasis on learning through earlier releases, while the open letter asks labs to pause a specific category of training until risks can be better managed.
Musk, an OpenAI cofounder who left the organization in 2018, has spoken about AI safety for years and has recently criticized OpenAI. Altman said he believed Musk was understandably stressed about artificial general intelligence safety, while also suggesting there could be other motivations. He said some of Musk’s behavior had hurt him, but wished Musk would look more closely at the work OpenAI was doing.
The letter’s proposal therefore sits within a wider disagreement over how AI safety should be pursued: by slowing training through a coordinated pause, or by continuing development while testing systems and allowing society to adapt. Its demand for public verification and participation from key actors reflects the challenge of making any pause meaningful across labs.