Background to the complaint

1. The European Innovation Council’s ‘Accelerator’ programme is a funding programme under Horizon Europe that offers support to start-ups and small and medium enterprises (SMEs) that have innovative products, services, or business models that could create new markets or disrupt existing ones in Europe and beyond.[1]

2. The complainant is a company, which applied for funds under the above programme in 2023, by submitting a proposal. In January 2024, the complainant was informed that its proposal had been rejected. The complainant did not appeal this decision within the prescribed 14-day deadline.

3. However, the complainant wrote to the European Innovation Council and SMEs Executive Agency (EISMEA)[2] in November 2024 and again in May 2025. It argued that, based on a technical opinion it had obtained in the meantime, one or more reviews of its proposal must have been prepared with the use of a third-party artificial intelligence (AI) system (such as an online chatbot). It also pointed out inconsistencies in the overall assessment of its proposal and said that its business secrets included in the proposal must have been disclosed to the third-party AI system, for which it requested compensation.

4. EISMEA replied to the complaint’s concerns in December 2024 and August 2025. It maintained its overall assessment of the proposal. Among other things, it said that it had contacted the three experts who had evaluated its proposal. Two of them confirmed the use of AI tools to perform only ancillary tasks, such as refining the language of the reviews and collecting information about markets for the complainant’s product. The third expert said he had not used any AI tool. The experts said that they had not disclosed any information submitted by the complainant to a third-party AI system. Also, EISMEA reminded the complainant that, under rules governing the funding programme at issue, no confidential information should be contained in the applications without clear indication of their restricted use.

5. Dissatisfied with EISMEA’s reply, the complainant turned to the Ombudsman in October 2025.

The inquiry

6. The Ombudsman found insufficient grounds to inquire into the substantive evaluation of the complainant’s proposal and the concerns regarding disclosure of sensitive information. The Ombudsman found EISMEA’s explanation of how the complainant’s proposal was evaluated to be clear and reasonable, including: that the limited use of AI in this case was acceptable; that the outcome of the complainant’s application depended on formal scores granted by the experts rather than their qualitative comments (which might have been prepared with some use of AI); that EISMEA could not second guess the scores made by independent expert evaluators; that proposals such as the one that the complainant submitted should not contain any confidential information without clear indication of their restricted use; and that the complainant could resubmit its proposal under the upcoming call.

7. Nonetheless, the Ombudsman opened an inquiry into the potentially systemic issue raised by this complaint, namely, the rules governing how expert evaluators contracted by the Commission, EISMEA, and other executive agencies in the context of similar EU funding programmes can use third-party AI systems when reviewing applications and proposals, as well as whether this use is transparent towards the applicants and duly recorded.

8. Although the complaint was brought against EISMEA, the Ombudsman decided to include the European Commission in the inquiry, as the latter adopted the rules at issue.

9. In the context of the inquiry, the Ombudsman inquiry team met with the representatives of the Commission and EISMEA,[3] and inspected relevant documents provided by these bodies. The report on the meeting was shared with the complainant, who did not provide comments on it.

Arguments presented to the Ombudsman

10. The complainant argued that EISMEA failed to ensure a fair, impartial and transparent evaluation of its proposal, as the evaluation relied heavily on AI-generated reviews rather than human assessment. In the complainant’s view, by allowing experts to delegate the review of proposals to AI, EISMEA failed to ensure the proper allocation and use of EU funds. In this regard, the complainant relied on a technical opinion it obtained according to which the AI must have been used in the evaluation of its proposal and not only in editing the reviews.

11. Also, the complainant contended that EISMEA’s handling of its concerns was superficial and dismissive, and that EISMEA did not engage with the substance of the arguments. Moreover, the only available appeal mechanism was limited by a 14-day deadline, which in the complainant’s view, was insufficient to obtain expert evidence concerning the use of AI in the review process at issue.

12. In its replies to the complainant, EISMEA said that the contracts for experts evaluating proposals submitted under Horizon Europe calls, such as the programme at issue, do not prevent experts from using AI tools to perform ancillary tasks, such as to improve the language and clarity of their reviews, as long as such use is in compliance with the contractual obligations to protect confidential information. EISMEA also said that the experts are briefed about the risks related to the use of AI and that they may not delegate the evaluation of proposals to AI tools.

13. During the meeting with the Ombudsman inquiry team, EISMEA representatives said that they are aware of concerns of applicants related to the use of third-party AI tools by expert evaluators, including the potential risk of disclosure of sensitive technical and business information, as well as personal data. Nevertheless, EISMEA acknowledged that third-party AI tools can be useful for certain ancillary tasks, including for language and style editing, which helps some experts to phrase their complex reviews more clearly. In this context, EISMEA instructs the reviewers on how to use AI tools responsibly. As a result, the quality of expert comments has overall improved, in particular in terms of language and style, which helps the applicants to better understand the reviews they obtain. Overall, EISMEA noted that, following the call in which the complainant took part in 2023, EISMEA received in 2024 only a few complaints regarding the use of third-party AI tools by external experts.

14. EISMEA said that the rules relevant to the use of AI by experts are contained in the model contract for experts and the annex thereto laying down a ‘Code of Conduct’ (Point 1.3.).[4] The Commission has developed these documents at ‘corporate level’, which means that they are applicable across EU programmes managed by the Commission, and not only to those managed by EISMEA. Also, EISMEA has continuously updated its own internal compulsory briefing for expert evaluators, which is based on the Commission’s standard briefing for Horizon Europe external expert evaluators, and which now contains specific instructions regarding the use of AI tools. In parallel, EISMEA has also updated its internal guidance for all units and programmes on how to handle AI-related complaints. The relevant documents were provided to the Ombudsman inquiry team for inspection.

15. The Commission confirmed that it continues to monitor developments in this area and, where necessary, adapts its guidance and documents to ensure that evaluation practices remain robust, fair, and trustworthy.

16. The Commission further noted that the Code of Conduct was updated on 15 December 2025 to clarify the conditions for the use of AI tools by experts. According to point 1.3. of the Code of Conduct, “Experts may not delegate the work to other persons or be replaced by other persons. AI tools may be used only for ancillary purposes and under the full responsibility of the expert; their use must be appropriately documented and made available to the contracting authority on request”.[5] The Commission clarified that there are many ways of preparing such documentation, but one way could be to keep two versions of the expert review, that is, one version before and another version after editing by AI tools. This would enable a comparison, if needed, as clear evidence that the initial version of the review was prepared by the expert without the use of AI tools. However, this form is not mandatory, and different ways of keeping the documentation are possible, which is discussed during the expert briefing.

17. The Commission added that other options were also under discussion, such as requiring external experts to declare the use of AI tools proactively, when submitting their reviews of proposals, and not only in case of subsequent complaints. However, it was considered that requiring declarations of any use of AI proactively would result in a need to record and verify a great number of additional documents. The competent administration would need to allocate to this additional task significant resources, considering that external experts participate in many different EU programmes managed not only by EISMEA, but also by the Commission, as well as by other executive agencies.

The Ombudsman's assessment

18. As regards the complainant’s concerns about the use of AI in assessing its proposal, the Ombudsman acknowledges that it is generally challenging, if possible at all, to establish with complete certainty, in a case like this one, whether and how external experts advising EU institutions or bodies have used AI to produce a review, opinion, or report. The Ombudsman understands that fully reliable methods detecting the use of AI are not yet available. In this specific case, EISMEA took reasonable and adequate measures to seek to address the complainant’s concerns. EISMEA reached out to the relevant experts to verify whether and how they used AI when reviewing the complainant’s proposal, and subsequently, it analysed the impact of potential issues on the overall outcome of the complainant’s application.

19. Given the challenges in detecting the use of AI, what can reasonably be expected from EU institutions and bodies in this regard is that they adopt clear rules for the use of AI by external experts and that they ensure that these rules are clearly communicated to contracted experts. In addition, EU institutions should ensure that the experts in turn properly document the use of AI tools.

20. In this context, the Ombudsman welcomes the Commission’s decision to update the Code of Conduct and lay down clear rules concerning the use of third-party AI by the experts for ancillary tasks only, while emphasising the experts’ responsibility for the substance of their reviews, as well as the need to document any use of third-party AI.[6] In the Ombudsman’s view, laying down such rules constitutes a systemic improvement.

21. Also, the Ombudsman takes note of the reasons for not prohibiting the use of third-party AI when reviewing proposals submitted in the context of EU funding programmes, namely, that the limited use of AI appears to be overall beneficial for the quality of submitted reviews. Moreover, as mentioned above, a ban in this regard could be difficult to enforce in practice. In this context, it is reasonable for the Commission to focus its efforts on training experts about the principles of responsible and traceable use of AI.[7]

22. In light of the above, the Ombudsman considers that no further inquiries into this complaint are justified.

23. That said, the Commission should seek to improve how experts’ use of AI is documented. Asking relevant experts for documentation of their use of AI only to react to a specific complaint is not optimal and may not always be sufficiently reliable. Not all experts, especially those reviewing multiple proposals or advising on many programmes in parallel, may succeed in keeping appropriate documentation, especially given that the form of this documentation is not specified by the Commission. If appropriate documentation is not created during the review process, experts contacted for information many months after the review at issue may have difficulty recalling its details. Also, if such experts are at the same time notified about a complaint brought against them, knowing this fact may dissuade them from disclosing accurate information, as doing so could expose them to contractual liability or other professional and reputational consequences.

24. Therefore, the Ombudsman suggests that the Commission explore further how experts advising on funding programmes could document their use of third-party AI during the review process and provide this information to the relevant EU funding bodies along with their reviews. Such records could focus on whether third-party AI was used at all, for which ancillary activity (drafting, background research, summarising documents etc.), and what was the AI’s specific contribution (such as researching information on the relevant market for the applicant’s product etc.).

25. The Ombudsman makes a corresponding suggestion for improvement below.

Conclusion

Based on the inquiry, the Ombudsman closes this case with the following conclusion:

As the Commission has recently specified that experts advising EU bodies in the context of funding programmes can use third-party AI systems for ancillary tasks only and must keep adequate documentation, no further inquiries into this complaint are justified.

The complainant, the Commission, and EISMEA will be informed of this decision.

Suggestions for improvement

The Commission should explore further how experts can document their use of AI during the review process and proactively provide this information along with their reviews to the relevant EU funding bodies.

Teresa Anjinho 
European Ombudsman 

Strasbourg, 28/10/2026 

 

[1] https://eic.ec.europa.eu/eic-funding-opportunities/eic-accelerator_en

[2] This executive agency is responsible for handling administrative matters concerning the EIC ‘Accelerator’ programme.

[3] https://www.ombudsman.europa.eu/en/doc/inspection-report/en/235911

[4] See the ‘Expert Model Contract’ at https://ec.europa.eu/info/funding-tenders/opportunities/docs/2021-2027/experts/model-contract_en.pdf. See the ‘Code of Conduct’ at https://ec.europa.eu/info/funding-tenders/opportunities/docs/2021-2027/experts/code-of-conduct_en.pdf.

[5] https://ec.europa.eu/info/funding-tenders/opportunities/docs/2021-2027/experts/code-of-conduct_en.pdf

[6] The Ombudsman has already held that, under the principles of good administration, EU institutions and bodies cannot delegate their tasks completely to AI systems, although they may use AI to support their activities and perform ancillary tasks. See Closing Note on the Strategic Initiative on how the European Commission decides on and uses artificial intelligence (SI/4/2024/MIK), https://www.ombudsman.europa.eu/en/doc/closing-note/en/196934.

[7] As also emphasised by the Ombudsman in the Closing Note, ibid.