UK launches Fundamental AI Research Lab: funding and compute to fix AI's core flaws
Published: 4 March 2026
The UK government is backing a new Fundamental AI Research Lab to push true breakthroughs in how AI works. Up to £40 million over six years is on the table, plus substantial in-kind access to large-scale compute through the national AI Research Resource.
The brief is clear: solve the stubborn problems holding AI back-hallucinations, short memory, and unpredictable reasoning-and explore new approaches that make systems more accurate, transparent, and trustworthy.
What's new
- Open call for ambitious, blue-sky AI research proposals from the UK's AI community.
- Funding up to £40m for fundamental work with long-term impact potential.
- Significant compute access to support training and evaluation at scale.
- Peer review panel chaired by Raia Hadsell (Google DeepMind VP, Research; DSIT AI Ambassador).
Why this matters to researchers
This lab isn't about scaling the same models with more data. It's about rethinking the foundations so AI can reliably support high-stakes domains. The goal: earlier diagnoses, more resilient infrastructure, faster scientific discovery, and better tools for people and public services.
Focus areas the call is primed to support
- Reducing hallucinations and error modes to improve reliability under real-world constraints.
- Long-term and compositional memory for persistent context and task continuity.
- Predictable reasoning and verifiable outputs for decision support and science.
- Transparency and interpretability so practitioners can understand and trust model behavior.
- New architectures and training paradigms that go beyond scaling existing systems.
Who should consider applying
- University teams and consortia tackling fundamental AI problems.
- Interdisciplinary groups combining AI, maths, computer science, engineering, and domain sciences.
- Researchers with high-risk/high-reward ideas that need compute and time to mature.
What reviewers will want to see
- A crisp problem statement tied to reliability, memory, reasoning, or transparency.
- A plausible path to empirical validation and comparison against strong baselines.
- Clear milestones, evaluation plans, and compute requirements.
- Potential to enable measurable benefits in healthcare, transport, science, or public services.
- Plans to share artifacts or insights that move the field forward.
Leadership and signals from government
Kanishka Narayan, AI Minister: "If we want this technology to be a force for good, we need to make sure the next big AI breakthroughs are made in Britain… If we are the ones breaking new ground on what AI can do, we can make sure our values are baked in from the outset."
Raia Hadsell, VP, Research at Google DeepMind and DSIT AI Ambassador: "Fundamental research that helps this technology achieve its full potential is key. The UK has the world-class talent and academic ecosystem to drive transformational research."
Dr Kedar Pandya, UKRI/EPSRC: "By backing ambitious, ground-breaking work, the new Fundamental AI Research Laboratory will unlock fresh capabilities, strengthen trust and reliability, and help the UK remain at the forefront of advancing AI for society and the economy."
Context: broader UKRI push
This call is an early step in the new UKRI AI Strategy, backed by £1.6 billion over four years to accelerate AI across UK science and research. Expect stronger investment in mathematics, computer science, and engineering-plus better access to tools, training, and infrastructure for teams across the UK.
Evidence the model can work
UKRI-backed AI is already delivering impact, from the RADAR system detecting real-time faults on the rail network, to the IXI Brain Atlas supporting dozens of clinical trials into degenerative diseases by improving brain scan analysis.
What to do next
- Scope a proposal centered on a core limitation of current AI systems, with a plan that advances the state of the art-not just scale.
- Define rigorous evaluation and a compute budget aligned to your method and milestones.
- Assemble a team that blends theory, engineering, and the target application domain where useful.
- Submit while the call is open; proposals will be peer reviewed under a panel chaired by Raia Hadsell.
For support and skills
Explore methods, tools, and training paths curated for scientists and research teams: AI for Science & Research
The opportunity is open now. If your work can make AI more reliable, transparent, and useful where it matters most, this is the time to pitch big.
Your membership also unlocks:
