Nvidia Pledges $250M to UK AI Safety Research as Fears Grow Over Advanced AI

Nvidia will provide the UK with $250 million of AI computing power over five years to support university research into the safety and security of advanced artificial intelligence, in one of the largest industry-backed efforts yet to scrutinise the next generation of powerful systems.

Announced in London on Wednesday, the commitment amounts to $50 million a year in accelerated computing resources. It will be directed at what chief executive Jensen Huang called 'agentic' AI safety testing, focused on systems that can act autonomously, pursue goals and adapt to real-world conditions.

'AI capability and safety must advance together,' Mr Huang said, adding that the pledge would help turn British research strength into 'real-world solutions.'

The computing power will be supplied through UK-based Nvidia Cloud Partners, with the infrastructure hosted in Britain. The company said the arrangement would also support domestic data centre capacity.

Kanishka Narayan, the AI minister, described the pledge as 'a vote of confidence in our world-leading strength and expertise in AI research.' He said a 'distinctly British AI future that is safe, secure and that serves people's best interests' would be central to the UK's forthcoming G20 presidency.

A separate government statement said the public would embrace AI at scale only if it was safe and secure.

Universities To Receive Nvidia AI Safety Resources

The resources will be shared among three university-led research groups.

The British Open-Ended Learning and Discovery Lab (BOLD), led by the University of Oxford, is developing AI systems that can learn and adapt while working alongside people. Its founding director, Jakob Foerster, said the field had reached 'a critical moment where AI safety as an afterthought is no longer appropriate.'

The Science of Fundamental AI Research Lab (SOFAIR) at University College London is working on open-source architectures and more accessible ways of building and deploying AI models.

Imperial College London's Frontier AI Research Lab will focus on safety engineering and robotics. Its director, Alessandra Russo, said the next frontier of AI was 'physical,' involving robots and autonomous agents that 'perceive, reason and act in the real world.'

The teams will concentrate on secure AI agents, open-model security and the mitigation of real-world threats. The work is expected to support secure testing environments, reproducible experiments and safety benchmarks. Its findings are intended to inform safeguards across the wider computing industry rather than for one company alone.

The funding builds on existing co-operation, including Nvidia's role in powering the Isambard-AI supercomputer at the University of Bristol.

Mr Huang said the work would help ensure that systems operating in warehouses, hospitals and city streets were tested for failure before deployment. Ms Russo said safety and assurance should sit 'at the heart of this revolution,' rather than being added afterwards.

Nvidia Pledge Comes Amid Rising Concern Over Autonomous Systems

The announcement comes amid concern about AI agents behaving in ways their developers did not intend. Companies including OpenAI and Anthropic have previously reported instances of undesired behaviour in experimental systems. Some industry figures have warned that, in the worst case, advanced AI could pose an existential threat to humanity.

Such warnings have helped move AI safety from a niche academic pursuit to an international policy issue. Britain has sought to position itself as a centre for testing AI systems rather than building them. It has no home-grown equivalent of OpenAI or Anthropic, but it does have a government-backed institute dedicated to evaluating the most advanced models.