UK research funders to allow AI in grant assessment

TL;DR:

• The Research Funding Policy Group, including UKRI and Wellcome, will update its joint position to let funders use generative AI when processing grant applications.
• Whole applications will not be fed into public AI tools, and all final funding decisions will remain with humans.
• Applicant disclosure of AI use is already running at 12–25% across major funders.

Eight of the UK’s biggest research funders are preparing to relax their stance on generative AI, clearing the way to use it in the assessment of grant applications. Speaking at the Association for Research Managers and Administrators conference, representatives from UK Research and Innovation and Wellcome said the Research Funding Policy Group would soon revise the cautious joint statement it issued in 2023.

From caution to qualified acceptance

The shift is notable given how firmly individual funders had drawn the line. Wellcome’s current policy rules out using AI to assess application quality or aid funding decisions, while UKRI bars reviewers and panellists from using the tools in assessment at all. Lydia Fulford, a senior funding policy manager at UKRI, said the updated statement would instead say funders “may use generative AI to process funding applications”, with the safeguard that no complete application is put into a publicly available tool and that humans take every final decision.

The driver is volume. Funders are grappling with surging application numbers, and UKRI has backed a University of Sheffield project exploring how large language models might support grant review. Some of that surge is itself attributed to applicants using AI to write bids — disclosure rates reached 25% for Wellcome early-career awards, 16% for British Heart Foundation grants and 12% at Cancer Research UK. Wellcome said it would not pass disclosure data to grant committees, to stop reviewers speculating about how much AI a bid involved.

Looking forward

For UK universities and researchers, the change formalises what many suspected was already creeping into practice. The tension now is one of trust and transparency: funders are asking applicants to declare AI use while preparing to deploy it themselves on the other side of the desk. With a disclosure “tick box” on UKRI’s platform already under review, the sector faces a familiar AI-governance question — how to capture efficiency gains in a high-stakes, fairness-sensitive process without eroding confidence that decisions are made on merit.