Safety & Reliability team
Safety & Reliability builds the evaluations, safeguards and monitoring that every model and feature passes through before and after release.
Evaluations
Measuring accuracy, honesty and misuse risk before anything ships.
Safeguards
Classifiers, policies and product design that prevent harmful use.
Factuality
Reducing confident mistakes and teaching models to say when they are unsure.
Featured
Policy · Nov 4, 2026
Core views on AI
Why we believe AI will change how the world works, why that makes reliability and safety urgent, and how those beliefs shape what ACATHRCA builds.
Safety · Nov 4, 2026
Publishing our usage policy
One clear policy now covers every ACATHRCA product. Here is what it says, why, and how we enforce it.
Publications
TitleCategoryDate
- Core views on AIWhy we believe AI will change how the world works, why that makes reliability and safety urgent, and how those beliefs shape what ACATHRCA builds.PolicyNov 4, 2026
- Publishing our usage policyOne clear policy now covers every ACATHRCA product. Here is what it says, why, and how we enforce it.SafetyNov 4, 2026
- How we think about responsible deploymentShipping AI to millions of people is a responsibility. These are the principles that guide what we release, when, and with which safeguards.PolicyJul 30, 2026
- Our research agendaFive questions guide ACATHRCA research, from better reasoning to cheaper inference. This is what we are working on and why.ResearchJul 15, 2026