Policy
Helping governments and institutions write AI rules that can be followed in practice.
Good AI policy needs two kinds of knowledge: what a society wants from the technology, and how the technology actually behaves. Drafts often have plenty of the first and too little of the second.
We bring the technical side into the room. We review draft rules against how AI systems are built and used, point out where a requirement cannot be measured or enforced, and suggest wording that can be.