Lead the development of the infrastructure that could define how society governs increasingly powerful AI systems.

As AI systems become increasingly capable, one question is becoming unavoidable: how can powerful systems be independently scrutinized when the organizations involved cannot fully share what they know?

AI developers are protecting frontier models worth billions of dollars. Auditors, safety researchers, and regulators often rely on sensitive evaluation methods, proprietary benchmarks, and restricted datasets that cannot be publicly disclosed. Yet meaningful oversight depends on both sides being able to collaborate.

At OpenMined, we're building the privacy-preserving infrastructure that makes trustworthy oversight possible without requiring trust. Using privacy-enhancing technologies, secure computation, and distributed systems, we're creating the rails that allow AI developers, auditors, safety institutes, researchers, and regulators to run evalu...