Tech giant commits $50 million annually in computing power to back Oxford, UCL, and Imperial College London research into secure autonomous AI
Nvidia has announced a $250 million five-year commitment to supply high-performance computing resources for AI safety and security research across the UK. The initiative will deliver $50 million in compute capacity each year to support key academic institutions, including the British Open-Ended Learning and Discovery Lab (BOLD) at the University of Oxford, the Science of Fundamental AI Research Lab (SOFAIR) at University College London (UCL), and Imperial College London.
Jensen Huang, founder and chief executive of Nvidia, highlighted that AI progress and safety engineering must move in lockstep. He noted that providing UK researchers with accelerated computing will allow them to build, evaluate, and reinforce safeguards for increasingly capable AI systems, converting academic excellence into practical solutions.
The compute allocation is designated specifically for research into agentic AI safety, security, and risk mitigation. Each participating institution brings distinct specialisations to the programme. BOLD, led by the University of Oxford, focuses on methodologies that enable AI systems to learn, adapt, and operate safely alongside humans. SOFAIR, led by UCL, investigates novel open-source AI architectures and accessible frameworks for building and deploying AI models. Meanwhile, Imperial College London drives progress in safety engineering protocols and robotics.
To reinforce the domestic technology ecosystem, Nvidia will fund compute access through UK-based cloud partners, hosting the infrastructure locally to establish a dedicated platform for British researchers. Details regarding provider selection, system architecture, and commercial arrangements will be finalised in due course.
The broader research roadmap encompasses secure autonomous agents, open-model security, and strategies to mitigate real-world threats. Investigators will utilise secure sandboxes, reproducible research environments, and safety evaluation benchmarks, leveraging Nvidia’s OpenShell secure runtime software and Sentry reference design. Findings from the initiative are intended to establish foundational safety standards for the wider global computing landscape.
