OpenAI publishes 'Governance of superintelligence'

OpenAI’s three most senior figures — chief executive Sam Altman, president Greg Brockman and chief scientist Ilya Sutskever — published a statement titled “Governance of superintelligence,” arguing that within the next ten years AI systems would likely exceed expert skill level in most domains and produce as much economic activity as one of today’s largest companies.

The essay proposed three elements. First, continued coordination among leading developers to make the transition to advanced AI safe and to smooth its integration into society, including voluntary coordination on the rate of frontier capability growth. Second, sustained investment in technical safety and alignment research proportional to the risk. Third, and most concretely, an international authority for the most capable systems, explicitly modelled on the International Atomic Energy Agency, empowered to inspect systems, require audits, test compliance with safety standards, and restrict degrees of deployment and security levels for the most powerful models.

The authors argued against a moratorium on capability development, writing that stopping the creation of superintelligence would effectively require “something like a global surveillance regime,” and that such a regime was not guaranteed to work even if attempted — a position read partly as a rebuttal to the Future of Life Institute’s open letter calling for a six-month pause, published two months earlier, which OpenAI’s leadership had not signed.

The essay drew a mixed response: some in the safety research community welcomed a frontier lab publicly endorsing binding international oversight of its own category of technology, while critics noted that a proposal to police the field’s largest, most capable models sat well with a company positioned to remain among that group, and that the IAEA comparison assumed a degree of state buy-in nowhere close to existing. It became a frequently cited reference point in the AI governance debate that intensified over the following two years, including at the UK’s Bletchley Park summit that November.