NVIDIA commits $250M in compute to UK agentic safety research

The commitment directs $250M of compute over five years to UK academic groups studying the safety and security of autonomous agents, at Oxford, UCL and Imperial.

It arrives in a week of agent-security news: Anthropic's worm warning, AWS's autonomous pentesting results and Perplexity's open-source sandbox. A hardware vendor funding safety research also helps position NVIDIA with UK policymakers.

NVIDIA did not specify how the compute will be allocated or which hardware generation it covers. Watch for funded projects and their publications.