Autonomous AI agents capable of writing and submitting grant applications without human involvement represent a fundamental threat to the integrity of research funding, according to experts speaking at a webinar organised by the League of European Research Universities.
Geraint Rees, vice provost for research, innovation and global engagement at UCL, drew a sharp distinction between generative AI tools, which help researchers polish applications, and agentic AI systems, which can independently gather information, draft proposals based on a researcher’s published work and funding criteria, and submit them autonomously. The marginal cost of producing an application, Rees warned, has effectively fallen to zero.
Research conducted by Rees and James Wilsdon, professor of research policy at UCL and executive director of the Research on Research Institute, found that grant application volumes across 12 funders in seven research systems rose 57 percent between the launch of ChatGPT in 2022 and the end of 2025, with 2026 data suggesting the pace is accelerating further.
The pair identified three compounding challenges: surging application volumes, quality compression as AI-polished proposals raise the floor without raising the ceiling, and long-term convergence, where agents writing and reviewing applications trained on identical data risk reducing funding decisions to a measure of how well AI simulates previously rewarded thinking.
Both researchers rejected banning AI as unenforceable, instead calling on funders and universities to redesign assessment frameworks around qualities that agentic AI cannot replicate, including career-stage research track records and demonstrable originality of ideas.
