NVIDIA commits $250m in computing access to UK AI safety research

Oxford-led BOLD, UCL-led SOFAIR and Imperial College London will receive support, while a separate Locai Labs collaboration will give developers one million synthetic UK personas

NVIDIA has committed computing access valued at $250 million over five years to support UK AI safety and security research.

NVIDIA has committed $50 million a year in computing access for five years to UK researchers working on the safety and security of AI agents. The majority of the $250 million program will support the wider research community, with a portion allocated to work involving university-led labs and Imperial College London.

Access will be funded through UK-based NVIDIA Cloud Partners using infrastructure hosted in the country. Provider selection, technical architecture and commercial terms remain to be agreed.

The announcement covers computing resources rather than cash grants to universities. Research will address secure AI agents, open-model security and ways to mitigate real-world threats, alongside controlled testing environments and reproducible safety benchmarks.

“AI capability and safety must advance together,” says Jensen Huang, Founder and CEO of NVIDIA. “The U.K. brings world-class scientific rigor to this essential engineering challenge.”

Oxford, UCL and Imperial take different research roles

The British Open-Ended Learning and Discovery Lab (BOLD), led by the University of Oxford, is developing approaches that help AI systems learn, adapt and operate alongside people.

UCL leads the Science of Fundamental AI Research Lab (SOFAIR), which is exploring open-source AI architectures and more accessible approaches to developing and deploying AI. Imperial College London’s work will include safety engineering and robotics.

For BOLD, the research includes how governance and human involvement can be built into systems as they develop.

“We are experiencing a critical moment where AI safety as an afterthought is no longer appropriate for the open-ended, evolving world,” says Jakob Foerster, Founding Director of BOLD. “This compute will help us explore transparent and human-centric approaches to AI that allow us to collectively rise to the challenge by building in governance and human-AI co-improvement from the ground up.”

The wider program will include research using NVIDIA OpenShell, its open, secure runtime software, and the NVIDIA Sentry reference design. Findings are intended to inform safeguards across the computing sector.

One million synthetic personas for testing UK AI

In a separate announcement, London-based Locai Labs has developed Nemotron-Personas-UK with NVIDIA, an open-source dataset designed to help developers train and evaluate AI against the UK’s demographic, regional and cultural diversity.

The dataset contains one million computer-generated personas covering 372 local authorities across England, Wales, Scotland and Northern Ireland. Locai Labs says the records are grounded in demographic, geographic and socioeconomic distributions, but no real person is represented and no personal or private data is used.

Proposed uses include testing model performance across education levels, age groups and rural and urban communities. Developers could also use the personas to simulate conversations, fine-tune models on culturally grounded data, and evaluate applications in healthcare, law and finance.

Differences between the four UK nations shaped the development process, including their census systems, educational qualifications and language distributions. Locai Labs says it also factored Celtic language proficiency into persona generation to support potential use in models working across UK languages.

The company positions the dataset as a way to improve representation and reduce bias. Its announcement does not provide results demonstrating a reduction in model bias.

Writing on LinkedIn, James Drayson, Founder and CEO of Locai Labs, described the aim as building “AI that better reflects the UK!”

UK infrastructure supports research and dataset development

Locai Labs produced the dataset using 18 NVIDIA graphics processing units across cloud and local computing resources, including its own Locai 1 AI computer. Drayson said on LinkedIn that the computers were designed in London and built in Greater Manchester.

NVIDIA’s research commitment also builds on existing university collaborations, including the University of Bristol’s Isambard-AI supercomputer, which uses NVIDIA Grace Hopper Superchips.

Nemotron-Personas-UK will be released under a Creative Commons Attribution 4.0 license. Locai Labs says the UK dataset will join the NVIDIA Nemotron Personas collection the week after its October 14 announcement, during GTC Berlin.