Focus area
How AI systems fail or get misused, and what that means where the models are bought, not built.
Models built elsewhere, used in African and Gulf markets.
Rules, buying and skills in government and business.
Fraud and weapons get cheaper. Firms skip safety to keep up.
Systems that fail quietly, or keep running when we want them off.
Nothing published here yet. See all our research.
Work by others that we build on. We are not affiliated with these authors.
The four main ways AI could go wrong.
Cases of AI systems that mislead people.
Open problems in making AI systems safe.
What governments and labs should do about extreme risks.
