Human oversight cannot scale to machine intelligence. Agents already act faster than the people meant to check them; at billions of agents, transacting and coordinating at machine speed, human verification is not inefficient but impossible.
It is our core belief that AI safety must be rethought from the ground up. We want to unlock the core deployments of AI in ways that advance our societies. Truly economical and societal mass deployment of AI can only happen when we can secure, trust, measure and transfer the risk of AI systems.
We are building the infrastructure for a future with fundamental, by-design guarantees on safety, trustworthiness and reliability.
If you believe that making AI truly safe is the most urgent challenge of our generation, that safety has to be done by design and not through post hoc and ad hoc evals, that safety is not a limiting force but an accelerant of beneficial mass deployment of AI, that safety has to be multidisciplinary by nature, and you have a demonstrably strong background in computer science, natural science or mathematics research.
Then reach out with a short introduction of who you are and what you are and have been working on.
This page is best viewed at 800 x 600.