ERC to add a Fundamentals of AI evaluation panel from its 2028 calls

The European Research Council (ERC) said on Tuesday 29 September 2026 that it will set up a new evaluation panel, Fundamentals of AI (PE12), "starting from the 2028 ERC calls under the EU’s future R&I framework programme". The reason it gives is AI's growth as a field: AI-related proposals represent "a significant share of submissions to the current Computer Science evaluation panel (PE6)".

What the ERC announced

The ERC press release is short, and these are its main points, in its own words where quoted:

• The panel: "a new thematic evaluation panel on Fundamentals of AI (PE12) will be established".
• When: "starting from the 2028 ERC calls under the EU’s future R&I framework programme".
• Why: AI "has developed into a major and rapidly expanding area of computer science", and the change is meant to "ensure that the growing volume of AI research proposals is evaluated in a dedicated scientific context".
• Scope: the panel "will cover fundamental and basic research aimed at advancing the foundations, methods, and theories of artificial intelligence".
• Topics listed: "machine learning, reasoning, intelligent and autonomous systems, computer vision, natural language processing, generative and multimodal AI, and its mathematical and statistical foundations."

The release describes AI proposals as a "significant share" of PE6 submissions. It does not give a number or percentage.

What changes for researchers applying to the ERC

The release draws one line that matters to applicants. PE12 is for work on AI itself, "rather than applications of AI in specific domains, which can be submitted to other panels." On our reading of that sentence, a project that uses AI as a tool in, say, a life-sciences or social-science question can still go to the panel for that field. A project whose aim is to advance AI's foundations, methods or theories is the kind the ERC says PE12 will cover.

The ERC's current panel list, published as Annex 1 ("Primary Panel Structure") of the ERC Work Programme 2027, lists 28 panels by our count: 11 in Physical Sciences and Engineering (PE1 to PE11), 9 in Life Sciences (LS1 to LS9) and 8 in Social Sciences and Humanities (SH1 to SH8). In that annex, PE6 Computer Science and Informatics is described as "Theoretical and experimental computer science, information processing, intelligent systems." On our reading, since the ERC says PE12 starts from the 2028 calls, applicants in the 2027 calls are still working with that structure.

What the release does not say

We read the ERC press release in full. It does not state:

• which ERC grant schemes the new panel will evaluate (for example Starting, Consolidator or Advanced Grants);
• a detailed panel description or keyword list in the format the Work Programme uses for existing panels, beyond the list of topics above;
• how PE6's own description, which currently includes "intelligent systems", will change once PE12 exists.

The release names no ERC Scientific Council decision document, and we did not find one linked from it. The panel list is Annex 1 of the ERC Work Programme 2027 (European Commission Decision C(2026)4907), which the ERC also publishes as a separate annex PDF.

Why it matters

On our reading, this is a structural signal: the ERC is treating fundamental AI research as large enough to need its own evaluators rather than sharing a panel with the rest of computer science.

On our reading, the boundary lines will matter to applicants. The current PE1 Mathematics panel already covers "mathematical foundations of computer science" and "statistics", and PE6 covers "intelligent systems", while PE12's listed scope includes "intelligent and autonomous systems" and AI's "mathematical and statistical foundations". The ERC release does not say how proposals near those edges will be allocated, so the detailed panel descriptions, when published, are the thing to watch.