# How the European Commission decides on and uses artificial intelligence 
- Author: European Ombudsman
- Date: 2024-03-15T00:00+01:00[Europe/Paris]
- [URL](https://www.ombudsman.europa.eu/en/opening-summary/en/183503)
---
|-------------------------------|
| President European Commission |

Dear President,

The rapid developments resulting from artificial intelligence (AI) will have enormous implications for public administrations. AI systems are already being developed or introduced to help public administrations improve the quality and efficiency of services they provide. In addition to tasks such as sorting and classification, translation and editing, and research, AI will also play a role in more complex decision making by analysing data or even suggesting solutions or policy options.

AI also poses major challenges, when it comes to the accuracy of results, potential bias, explainability and human oversight not least when it is or will be used in decision-making processes that affect the public directly.

The manner in which the EU administration introduces AI should therefore be transparent and allow for sufficient scrutiny. The public should be able verify that the administration has done proper due diligence, that decisions to deploy AI involved all the relevant actors, and that procedures are in place to review the use of these AI systems based on experience. It is also crucial that EU institutions make sure that AI supports their tasks but do not replace them in making decisions affecting citizens. To date, these are the types of issues that members of the European Network of Ombudsmen have encountered when examining the use of AI at the national level.

Against this background, I have decided to open a strategic initiative to collect information about how the European Commission, as the largest EU institution, decides on and applies AI, as well as similar advanced information technologies, in its work. The initiative also aims to raise awareness about the need to ensure accountability in how decisions to introduce this technology are taken and subsequently implemented and monitored.

The European Commission recently published a communication on 'Artificial Intelligence in the European Commission'.[\[1\]](#_ftn1){#_ftnref1} In it, the Commission has shared important information on AI systems that it already deploys. It has also shared information on AI systems in development and testing. I very much welcome this first step towards increasing the information available to the public.

In my view, three systems described by the Commission in general terms are of particular importance to the right to good administration:

**- Analysis of feedback from the public** [\[2\]](#_ftn2){#_ftnref2}: It seems that this tool may give EU citizens the possibly to convey their views more effectively to the Commission on envisaged policy proposals, which I welcome. At the same time, I would like to learn more about how the Commission assesses and plans to address potential risks resulting from the deployment of tools of this kind, such as the possibility for some actors to manipulate the perception of support in the EU for specific policy options.

**- Competition case management** [**\[3\]**](#_ftn3){#_ftnref3}: This tool is geared at improving the Commission's ability to uncover infringements of competition rules having the greatest impact on the EU internal market. At the same time, the Commission will need to protect its investigations against any error, bias, as well as the procedural rights of the persons concerned.

**-** **Complaint handling** [**\[4\]**](#_ftn4){#_ftnref4}: As you know, my Office handles complaints concerning how the Commission itself deals with infringement and other type of complaints. I would be keen to know how the Commission intends to ensure protection against 'automation bias', as a result of which its complaint handlers could disproportionately trust solutions proposed by AI. Each complaint should be handled with sufficient care by human decision makers and possible past mistakes should not be multiplied due to automation.

The Commission already has significant experience in automation and is planning to considerably expand the use of AI tools in its work. I would appreciate if, in reply to this letter, you could describe in more detail the functionality of the tools concerning the analysis of feedback from the public, competition case management and complaint handling, as currently envisaged. It would also be useful to receive information about how the Commission intends to ensure the transparency and accountability of the processes for introducing and monitoring the use of these appliations. More specific questions are listed in the Annex.

I would be grateful to receive the Commission's reply to my questions by **14 June 2024**. I would also like to organise a meeting between my inquiry team and Commission representatives to address some additional questions regarding the above issues. I would like this meeting to take place in June or July 2024.

My Office has been in contact with the European Data Protection Supervisor in advance of launching this strategic initiative, given their specific role in this area, to seek to avoid overlap.

Please also note that, as one of the objectives of this strategic initiative is to enhance transparency of how the EU administration uses or plans to use AI in its work and decision making, I intend to publish your reply. However, if you consider some information to be confidential, please clearly mark it as such and, if possible, include it in a separate annex.[^\[5\]^](#_ftn5){#_ftnref5}

Should you require any further information or clarifications on this initiative, please contact the responsible inquiries officer, Mr Michał Krajewski.

Yours sincerely,

Emily O'Reilly  
European Ombudsman

Strasbourg, 15/03/2024

Annex
-----

***Technical functionalities and the automation of tasks***

1) What will the specific functionalities of the tools envisaged by the Commission to analyse the feedback from the public, manage competition cases and support complaint handling be? What specific tasks will these tools perform and how they will support decision making by humans?

2) What general risks can result from the use of these tools to the right to good administration and how can these risks be addressed?[\[6\]](#_ftn6){#_ftnref6}

***Decision making on the use of AI***

3) How will the Commission decide on the use of these tools? By which internal body will decisions be taken (will they involve the College of Commissioners, for example)? How will the Commission decide on the procedure for approving the use of these tools?

***Transparency of how decisions on the use of AI are taken***

4) What information about these tools has already been made available? Does the Commission intend to make further information publically available?

5) Has the Commission consulted or will it consult any other EU institutions or bodies or stakeholders outside the EU administration (for example, interest representing groups, civil society, wider public consultation) on the introduction or further development of these tools?

***Accountability regarding the use of AI tools***

6) How transparent and explainable will the use of these tools in general and in individual cases be?

7) How exactly do human decision makers verify the output of these tools?

8) Will the Commission regularly monitor the practical application of these tools and which entities within the Commission will oversee this?

[\[1\]](#_ftnref1){#_ftn1} Artificial Intelligence in the European Commission (AI@EC): A strategic vision to foster the development and use of lawful, safe and trustworthy Artificial Intelligence systems in the European Commission: <https://commission.europa.eu/system/files/2024-01/EN%20Artificial%20Intelligence%20in%20the%20European%20Commission.PDF>

[\[2\]](#_ftnref2){#_ftn2} Page 8: *"Analysis of feedback from the public: the Commission systematically consults the public on new legislative proposals (public consultation via the 'Have Your Say' portal). The feedback provided by the public is analysed by the Commission to help develop the legislative proposals."*

[\[3\]](#_ftnref3){#_ftn3} Page 8: "*Competition investigations rely heavily on the analysis of vast volumes of documents and other data, to extract valuable insights and gather evidence. AI facilitates their efficient search and analysis, enhancing the investigative process. Additionally, it is particularly important for ex officio investigations and soon for the enforcing* \[of\]*the Digital Markets Act (DMA).*"

[\[4\]](#_ftnref4){#_ftn4} Page 9: *"Complaints Handling with AI (CHAI): the Commission is exploring generative AI and Large Language Models (LLM) technologies to help case handlers dealing with complaints in shorter timespans by providing 'smart search' semantic search capabilities and 'smart drafting' allowing for re-use of past replies from similar complaints capabilities."*

[\[5\]](#_ftnref5){#_ftn5} Encrypted emails can be sent to our dedicated mailbox.Information and documents of this kind will be deleted from the European Ombudsman's files shortly after the strategic initiative has ended.

[\[6\]](#_ftnref6){#_ftn6} This question does ++not++ relate to risks involved by the processing of personal data. This question relates to the elements of good administration, such as the duty of careful and impartial examination of all relevant information in each case and to state reasons for any decisions.