How to Govern AI in a World in Rupture

Without law, AI answers to no one. It decides who gets a loan, a job, or a medical treatment. It picks which family is turned back at the border, whose face the cameras track, what neighborhood the police are sent to. AI can turn any of them down over a name, an address, and the color of their skin. And no one is accountable for the result. Goodwill and self-regulation will not fix this. Only law will. 

The challenge in establishing these laws is that they cannot apply only to the AI we know today. It must address the evolutions of AI still to come, such as autonomous agents. It has to protect people, whatever the technology. 

Meanwhile, what humans need from the law is simple. People should have the right to know when a machine is deciding their fate, to understand why, and to challenge the decision. The technology will keep changing. These rights can’t. 

The good news is that the hard part of this work is already done. Two years ago, a group of countries wrote down these rules. The result was the Council of Europe Framework Convention on AI, the first of its kind. This binding international treaty commits states to AI systems that respect human rights, democracy and the rule of law. With 21 signatories, including all of the G7, and a first ratification by the European Union in May 2026, it offers the common base the world gathering at the AI for Good Global Summit in Geneva is looking for.