AI triage quietly rejected grant proposals, U.K. funder admits

A major United Kingdom (U.K.) research funder used an artificial intelligence (AI) system to screen and reject grant proposals, according to a report in the journal Science [256067]. The system, described as "AI triage," sorted applications before they reached human reviewers, and proposals flagged by the AI were turned away without full peer review [256067]. The funder has not publicly detailed how many proposals were rejected this way or what criteria the AI applied [256067].

The disclosure raises questions about fairness and transparency in research funding [256067]. Peer review is the standard method for judging scientific work, and researchers expect their proposals to be assessed by human experts [256067]. Critics worry that AI screening may introduce hidden bias or filter out unconventional ideas [256067]. The funder says the tool was meant to manage a growing number of applications [256067]. Details of the system and its impact remain limited [256067].

The case highlights a broader concern about AI's growing role in science. Experts warn that AI may weaken the independence of research in ways that are hard to see [256001]. Science depends on independent thinking, and researchers must question results, test claims, and avoid outside pressure [256001]. AI can blur these safeguards by repeating biases hidden in its training data or pushing researchers toward popular answers instead of true ones [256001]. The risk is not that AI replaces scientists, but that it slowly reduces their freedom to think independently [256001].

Researchers and experts say AI-augmented science must remain contestable, meaning its methods, data, and conclusions must stay open to question, testing, and challenge by other researchers [256074]. If AI-driven results cannot be checked or disputed, trust in science weakens [256074]. Clear rules, transparency, and human judgment are needed to protect the integrity of research [256001][256074].