Charter, Control, Compute, Constitution: What Governs AI?

Governments across the globe are trying to find solutions and draft legislation. Yet, almost none of the decisions that actually determine who controls artificial intelligence are being made in parliaments. They are being made in corporate structures. When OpenAI restructured in 2025, its governance was settled not by legislation, but by two state attorneys general, a share class, and a commercial counterparty with the beneficiary of the original charter defined, in Delaware's own words, as "all of humanity". When a government asks whether its data is sovereign, the answer turns on which corporate group can be compelled, not on where the servers sit. As Australia prepares to legislate standards for AI and data centres, it faces the same choice: regulate the asset, technology, application, or regulate the entity.

And there is now a second constitution in play. The leading labs have begun publishing constitutions for their models. These documents set out values, priorities, and limits, drafted unilaterally, binding no one, and yet governing systems used by hundreds of millions of people that directly affect their thinking and decision-making.

In this public lecture organised jointly by the RMIT University School of Law and CoBL Tech, as well as the Latrobe Law School, Professor Andhov asks what it means that corporate law has become the operative law of AI, and who, if anyone, has standing under either constitution.