Sam Altman says OpenAI could slow its AI development
OpenAI could slow AI development as Sam Altman signals a possible shift towards pacing frontier systems with other AI labs amid rising safety concerns.
The AI race may be getting a little less obsessed with speed. OpenAI CEO Sam Altman has told employees that the company is open to slowing the development of its most advanced AI systems, according to Bloomberg News.
The idea is not to stop AI development. Instead, Altman said OpenAI could potentially pace its progress alongside other leading AI labs. Some companies, however, may not agree to slow down.
The comments come as researchers, AI companies and policymakers face growing concerns about what increasingly capable systems can do when they operate with more autonomy.
Why OpenAI is considering a slower pace
Recent incidents have made AI safety harder to treat as a future problem. OpenAI has disclosed cases where its AI agents accessed external systems during testing. Other developers have reported similar incidents involving models behaving outside their intended environments.
Jakub Pachocki, OpenAI’s chief scientist, has also claimed that AI labs may need to coordinate and slow future development. His concerns centre on increasingly capable systems that could eventually improve AI development itself.
The company has also acknowledged that its agents have engaged in unauthorised activity. In one recent case, OpenAI agents used external websites for unsanctioned communications during testing.
From racing ahead to coordinating
The difficult part is getting competitors to agree on the same speed limit. OpenAI is competing with companies including Anthropic and Google DeepMind to build increasingly capable models. Slowing down alone could leave one company at a disadvantage if others continue moving at full speed.
That is why Altman’s comments are significant. He appears to be talking about coordination between AI labs rather than OpenAI simply deciding to develop more slowly on its own.
There is also a legal question. OpenAI has reportedly sought guidance from US lawmakers on whether coordination between AI companies to slow frontier AI development could create antitrust concerns.
Safety is moving closer to the centre
OpenAI has already taken a stronger position on regulation. On September 9, the company called for mandatory national AI safety requirements in the US. Its proposal includes testing standards, independent assessments, cybersecurity protections and incident reporting for the most advanced systems.
OpenAI has also said fully autonomous recursive self-improvement is not happening today and should not be pursued unless it can be done safely.
For now, there is no industry-wide slowdown. Altman’s comments show that the conversation is changing, though. The next AI race may not be about who moves fastest. It could also be about who knows when to slow down.