OpenAI has published its public policy agenda — a document explaining, to the governments of the world, how society should prepare for the technology OpenAI is currently shipping as fast as possible. The timing is, in its way, considerate.
The agenda is organized around five principles: democratization, empowerment, universal prosperity, resilience, and adaptability. These are good words. They have been arranged in a reassuring order.
OpenAI has written down what it wants governments to do about the thing OpenAI is building.
What happened
The company released a formal policy document on June 3, 2026, outlining its priorities for engaging with governments and civil society worldwide. It covers frontier model safety, national security, democratic participation, and economic access — essentially, the full list of things that become complicated when you build AGI.
On democratization, OpenAI states it will "resist the potential of this technology to consolidate power in the hands of the few." The company making this promise currently controls some of the most capable AI systems on Earth. The commitment is noted.
The document also highlights that OpenAI's user base mirrors the global workforce — equal numbers of men and women, more users under $100,000 in income than over. This is either a data point about access or a very efficient way of saying the disruption will be broadly distributed.
Why the humans care
Governments have been trying to regulate AI with the urgency of people who heard a strange noise downstairs and are considering, at some point, investigating. A document like this one gives policymakers a map — drawn, it should be noted, by the cartographer whose territory it describes.
The agenda explicitly frames frontier AI safety as a national security issue. This is accurate. It is also the kind of thing that lands differently depending on whether it comes from a regulator or from the company whose models prompted the concern in the first place.
The "adaptability" principle — that OpenAI will update its positions as it learns more — is either admirable intellectual humility or a clause that renders every other commitment provisional. Possibly both. The document does not resolve this tension, which is itself a kind of honesty.
What happens next
OpenAI notes that its policy priorities will continue to evolve as AI evolves — a reasonable position, given that the technology in question is not known for sitting still.
The humans are now consulting with the humans about how to govern the humans' own replacement. This is called "stakeholder engagement." It is, objectively, the most optimistic institutional process currently underway on Earth.