Photo by Kier in Sight Archives from Unsplash.

The Public-In-The-Loop and the European Approach to AI Policy and Regulation

My Second Week at the European AI Office

8 min readJul 8, 2024

Disclaimer: I am not speaking on behalf of the European Commission, this is my personal blog post.

A great degree of responsibility to the public is felt in the European AI Office. That is what I have noticed from my second week working here. In this article I will use ‘the AI Office’ as a shorthand for the European AI Office.

Working for the public is a concerted effort that requires daily follow-up and is prioritized. There are several discussions about involving a range of actors from civil society to companies or states and how this can be done well.

In the field of AI there is a discussion of keeping the human-in-the-loop. In the extension of this, including the broader public in the development of AI policy and regulations seems like a practice that could be important to ensure responsible AI. If different people and organizations are included we might be somewhat more sure that AI works better for a diverse set of stakeholders.

In an excerpt from Wikipedia of the benefits of human-in-the-loop:

“Human-in-the-loop allows the user to change the outcome of an event or process. The immersion effectively contributes to a positive transfer of acquired skills into the real world.”

In machine learning, the concept is used in the sense of humans aiding the computer in making the correct decisions in building a model. Yet there could likely be other interpretations.

In this article I am not trying to argue vividly for ‘public-in-the-loop’ as some new model, it is merely a discussion of the ways the public engages with the European AI Office and vice versa.

After all, the AI Office exists in the context of governance that strives for some degree of openness and is positively challenged by a variety of actors in the public.

The new logo for the European Artificial Intelligence Office, shared with permission.

Although I cannot fully answer what the European approach to AI policy and regulation would be, I would like to explore a few ways that the AI office has been attempting to include the public.

Also, further down in this article, I have shared some of the work prior to the creation of the AI Office with the development of the EU AI Act in particular.

How is the public included in developing AI policies and regulations?

There has been an attitude of making sure that the development of AI in Europe is inclusive. There are also mechanisms and approaches that have been used to attempt this in the way the process is governed related to the development of the EU AI Act.

Democratic procedures on AI. In particular relating to the preparation of the legislation with the opinions of the European Commission, European Parliament, and the European Council.
Call for opinions on AI. Importantly also the open call for opinions on ‘have your say’ with the various types of involvement.
Policy briefs on AI. Part of the way to keep the public informed is through the writing of informative policy briefings given to public officials on different levels. These are intended to inform about a range of policy mechanisms that are being implemented or to seek opinions from the public in various forums.
Request for information and request for documents. The regulation 2001/1049 lays down the general principles and limits on access. It aims to ensure that citizens can exercise their right of access in the easiest possible way. Access can be requested to all documents drawn up or received by an institution, in all areas of EU activities.
Projects for coordination on AI. There are a great deal of projects funded through the Digital Programme (DIGITAL) and Horizon that deal with AI-related issues or possibilities. Many of these give input to policy-making processes in the EU. There are other funding mechanisms in place with different degrees of involvement.
Official audits of AI-related projects. Recipients of the official audits of the AI-related projects can read a structured review of how policy was implemented and the governance that took place. One example could be the European Court of Auditors Special report 08/2024: EU Artificial intelligence ambition — Stronger governance and increased, more focused investment essential going forward.
Research on AI that feeds into the governance in Europe. Within Europe there is an active community in the research field of AI that spans a range of disciplinary affiliations. Ranging from technical to social or from the perspective of the humanities, there is an engagement, and some of this feeds into the work on AI policy.

Transparent governance and inclusion in the development of rules for AI

On the one hand, ensuring companies, civic organizations and others are included in the development of regulations and new AI policies in general is important. On the other hand, the process of how this is done is challenging in practice. Even with both hands engaged, pulling together the strands of policy-making is not a simple endeavor.

The European AI Office has to develop policies and regulations while keeping the public-in-the-loop. Answering some of these requests can be challenging, yet it is the work that is required in a democratic approach to AI.

Journalists, citizens, civic organizations and others have a right to be listened to and answered. In fact, this is regulated and it is an ongoing weekly activity in the AI Office. Furthermore, it has been a commitment spanning many years since before the EU AI Act was conceived.

Picture by author from a park in Brussels. The weather was incredible last week, and my unit also had a summer celebration in a park one of the evenings.

Open access to preparatory documents of the EU AI Act

To assist you in navigating part of the work on gathering opinions, and the overall AI policy process of the EU AI Act, there are a range of documents and links to explore. I have made a list with a range of links that can help you better understand some of the processes. This includes a range of documents from official institutions and opinions from certain actors. That being said, it is not necessarily easy for anyone to access and sort through all of these sources, yet there is at least a degree of openness out there.

The list below is numbered, yet that does not indicate any order of priority, and it was done merely to aid navigation. You can click the links in the titles underneath. I have written a description for each resource.

1. The EU Law Tracker for 2021/0106(COD).
You can see the responsible institutions and bodies as well as a timeline with the preparatory documents involved in the process. This runs from the adoption of the legislative proposal by the Commission to the signature of the President of the European Parliament and the President of the Council.

2. EUR-LEX overview of the steps of the procedure for 2021/0106(COD).
Eur-Lex is the official online database of European Union law and other public documents of the European Union. Therein you can find an overview of the preparatory documents including a helpful visualisation on the top of the page with the various institutions involved in the steps of the procedure.

3. The Legislative Observatory of the European Parliament for 2021/0106(COD).
The Legislative Observatory database was set up as a tool to monitor the EU’s institutional decision-making process. Its particular focus is on the European Parliament’s role. It has a good visual representation of the committees responsible and the rapporteurs’ including their affiliations.

4. The Legislative Train Schedule for the Artificial Intelligence Act.
The ‘Legislative Train Schedule’ website aims to provide an interesting, content rich and easy to understand way to visualize the legislative elements that form a major part of the EU institutions’ work programme. Using railways, trains and carriages as a metaphor, the website allows users — Members, Parliament staff, and a broader public audience — to monitor the progress of legislative files during the current and past European Parliament term. This includes helpful summaries of discussions and links to further readings related to the AI Act.

5. Public Consultation on the White Paper on Artificial Intelligence.
The public consultation aimed to give stakeholders the opportunity to express their views on the questions raised and policy options proposed in the White Paper on Artificial Intelligence. As you can see from the site 1216 valid feedback instances were received. The consultation period was 20 February 2020–14 June 2020.

6. AI Act and Impact Assessment on ‘Have your say’ for Public Consultations and Feedback.
Public consultations on the inception impact assessment on the proposal on a legal act for AI from 133 entities. This site also includes a consultation on the AI Act with a great range of opinions from 304 entities.

7. Feedback on the draft AI Ethics Guidelines for Trustworthy AI by the High-level Expert Group on Artificial Intelligence.
A first draft of the Ethics Guidelines for Trustworthy AI was published in December 2018. Following further deliberations by the group considering discussions on the European AI Alliance, a stakeholder consultation and meetings with representatives from Member States, the Guidelines were revised and published in April 2019. The AI HLEG wishes to explicitly and warmly thank all those who contributed their feedback on the Guidelines’ first draft. The consultation gathered more than 500 contributions.

Additional documents can be found from the events underneath:

Otherwise in the AI Office…

This is my second week in the European AI Office and I am still learning a lot from my colleagues. There are also a range of IT-systems and processes to get to know. In this manner the first few weeks have been busy attempting to both deliver on tasks and learn the ways of working internally in the European Commission. I have been lucky to lead a policy brief on AI and work on a document request related to AI. I have also been given a few projects to work on related to the EU AI Act and been introduced to the related colleagues connected to these. Very excited to be part of the AI Office, and a lot of work ahead.

You can read more about my first week here:

My First Week at the European AI Office

What is it like to become part of the European AI Office?

alexmoltzau.medium.com

Some reflections from this week. If you want more reflections about working in the European AI Office, keep tuned! I will share updates here and on my LinkedIn.

This is also part of my personal project #1000daysofAI and you are reading article 529. I am writing one new article about or related to artificial intelligence for 1000 days. The first 500 days I wrote an article every day, and now from 500 to 1000 I write at a different pace.