The UK government is launching a new research initiative aimed at tackling one of the most persistent and frustrating problems in modern AI: hallucinations.
Hallucinations occur when models confidently spit out false or fabricated information. Hallucinations remain a major barrier to deploying generative AI in high-stakes domains like medicine, law, and finance, where fabricated outputs can cause real harm.
Recent evaluations of even the most advanced models show persistent issues, with some reasoning-focused systems hallucinating more frequently on complex tasks.
On Wednesday (4 March), the Department for Science, Innovation and Technology (DSIT) announced up to £40 million in funding over six years for the Fundamental AI Research Lab, a new facility dedicated to “blue-sky”- high-risk, high-reward – research into the core limitations of today’s AI systems.
The lab will focus on longstanding flaws that continue to undermine trust in AI, including:
- Hallucinations (generating plausible but incorrect outputs)
- Unreliable or short memory in models
- Unpredictable reasoning that can lead to erratic behavior
Rather than simply scaling up existing large language models with more data and compute, the dominant approach in the US and China, the initiative aims to support bold, fundamental work to rethink how AI is built from the ground up.
The goal is to create more accurate, transparent, and trustworthy systems capable of transforming sectors like healthcare (earlier disease diagnoses), transport (resilient infrastructure), science (faster discoveries), and everyday public services.
The funding comes alongside substantial in-kind access to the UK’s AI Research Resource compute capacity, valued in the tens of millions of pounds, giving researchers the horsepower needed for cutting-edge experiments.
“AI is already doing things we could never have imagined just a few years ago, like helping to diagnose cancer,” said AI Minister Kanishka Narayan.
“It can and will do even more, but 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. This is a critical part of our mission to make AI work for everyone.”
The lab builds directly on the UK’s recently unveiled UKRI AI Strategy, which commits a record £1.6 billion over the next four years to position Britain as a global leader in AI for science and research.
It also complements existing efforts like the AI Security Institute (AISI), which focuses on frontier model testing and safety.
Leave a Reply