Arma 2026: Stance on generative AI softens, but final funding decisions still taken “by humans”
Research funders in the UK are clarifying their position on generative artificial intelligence to allow its use in the processing of grant applications, members of a group of major funders have said.
At the Association for Research Managers and Administrators (Arma) conference on 18 June, representatives from UK Research and Innovation and Wellcome said the Research Funding Policy Group—which includes UKRI and Wellcome alongside six other major funders—will soon update their joint position on the use of generative AI tools in funding applications and assessment.
The group released a statement in 2023 which said the use of generative AI “presents potential risks for research”, while individual funders have gone further in their own policies. Wellcome’s policy says “we do not use generative AI tools to assess the quality of grant applications or to aid funding decisions”, while UKRI’s policy says that assessors, including reviewers and panelists, “must not use generative AI tools as part of their assessment activities”.
Shift in stance
But speaking at Arma, Lydia Fulford, senior funding policy manager at UKRI, said the joint statement would be updated such that funders “may use generative AI to process funding applications” but that, “as ever, we will not put a whole application into a publicly available generative AI tool”.
Fulford also said the clarified statement will “include that all final funding decisions will be undertaken by humans”.
The shift in position by the Research Funding Policy Group—which also includes the Association of Medical Research Charities, the British Heart Foundation, Cancer Research UK, the National Institute for Health and Care Research, the Royal Academy of Engineering and the Royal Society—comes as funders grapple with skyrocketing volumes of grant applications.
There are suggestions that to tackle such growth in applications, funders could turn to AI, and UKRI has funded a project led by the University of Sheffield to explore how large language models can be used in the grant review process.
Disclosure rates revealed
Some have suggested that researchers’ use of AI in putting together grant applications is a key driver in the rise in volume seen by funders.
Fulford stressed that trust was the “most important thing” around the use of AI, both by applicants and funders, but especially around the disclosure of AI use in applications, presenting applicant disclosure data for three different funders.
Some 25 per cent of applicants for Wellcome’s early career awards last January said they had used AI, alongside 16 per cent of applicants for British Heart Foundation grants in 2026, and 12 per cent of applicants for Cancer Research UK grants.
Sue Russell, senior policy funding adviser at Wellcome, said that such disclosure levels were “quite high” but that the funder would not pass on which applicants had used AI to its grants committees to avoid reviewers speculating as to how much AI tools had been used.
A research manager at the Arma conference also raised concerns that there is currently a “box to tick” within UKRI’s Funding Service platform to allow applicants to disclose their AI use, and Fulford said this was an issue the funder is “taking forward and looking at”.
