OpenAI’s leaders say the speed of AI development may outstrip the ability of existing authorities to manage its risks. They proposed an international organization to coordinate oversight of the most capable systems, while leaving open how such a body would work and what authority it would have.
Why OpenAI wants international coordination
In a company blog post, founder Sam Altman, President Greg Brockman and Chief Scientist Ilya Sutskever argued that advanced AI development needs coordination among leading efforts. They said this could help maintain safety and make the integration of these systems into society smoother.
The proposal focuses on systems that pass a certain threshold of capability or use of resources such as compute. Those efforts, the leaders wrote, could fall under an international authority able to inspect systems, require audits, test compliance with safety standards, and set limits on deployment and security.
The idea is not that AI can be managed by companies alone. The post recognizes that the technology’s pace creates a challenge for existing authorities, and that some shared oversight may be needed as systems become more capable.
A model drawn from nuclear oversight
OpenAI’s leaders pointed to the International Atomic Energy Agency as a possible model. The agency is the UN’s official body for international collaboration on nuclear power issues.
An AI oversight organization modeled on it might not be able to stop a bad actor directly. It could still help establish international standards and track whether countries and developers follow shared agreements. That would provide a starting point for governance, even if it could not guarantee compliance on its own.
The proposal remains broad. The post calls for an authority with tools to inspect, audit and test, but does not set out a finished structure or a detailed plan for how those powers would be applied.
Compute and energy as measurable signals
One practical suggestion is to monitor the compute power and energy used for AI research. OpenAI’s leaders described these as among the relatively few objective measures that could be reported and tracked.
Measuring resources would not answer every question about how AI should be used. But it could give oversight bodies a way to observe the scale of research and development, then audit it in a manner similar to other industries. The company suggested that smaller companies might be exempt to avoid stifling early innovation.
This approach places attention on resources that can be counted, rather than relying only on broad judgments about whether a particular use of AI is acceptable. It could help authorities identify large development efforts for closer scrutiny, though the post does not specify the threshold for oversight.
Support for regulation, with open questions
AI researcher and critic Timnit Gebru made a related argument in an interview with The Guardian, saying: “Unless there is external pressure to do something different, companies are not just going to self-regulate. We need regulation and we need something better than just a profit motive.”
OpenAI’s call signals support from a leading AI company for governance beyond public hearings. Yet the company’s leaders also conceded that they do not know how to design the oversight mechanism. The proposal starts a conversation, but leaves core questions about authority, standards and enforcement unresolved.
They also expressed support for slowing AI development, but described no immediate plan to do so. Their stated reasons include the technology’s potential to improve society and the possibility that bad actors could continue advancing quickly. That tension captures the challenge behind the proposal: building safeguards while development continues.