European Commission proposes the AI Act

Background

In the European Union, law-making is split between institutions. The European Commission, the EU's executive body, holds the right to put forward legislative proposals. The Council of the European Union, where member-state governments sit, and the directly elected European Parliament then examine that text, and a law can only take effect once the two agree on a common version. A Commission proposal is therefore the opening move in a process, not the end of one.

The proposal

In April 2021 the European Commission submitted its proposal for an Artificial Intelligence Act ^[1]. In the Council's list of documents on the file, the proposal is dated 14 April 2021 ^[1]. It was submitted by Thierry Breton, the Commissioner for Internal Market — the member of the Commission whose portfolio is the internal market ^[1].

A proposal of this kind is a draft: it sets out what the Commission thinks the rules should be, and it is then read, questioned and amended by the other two institutions. That is why the document's date and authorship matter for following the file — they mark the point at which the text became an official object of negotiation rather than an idea inside the Commission.

Discussion in the Council

In October 2021 the Council discussed the proposed act ^[2]. The discussion weighed the societal and economic benefits of AI against the need to protect privacy and to ensure safety and security ^[2].

That pairing captures the tension built into the file. AI systems can be useful to economies and to public services, but they also depend on data, and they can affect people's safety and their private lives. A council of national governments discussing such a proposal is essentially asking how much room to give the technology and how much protection to require around it, and where the balance should sit when the two pull in different directions.

The Council's position

On 6 December 2022 the Council adopted its position on the new rules for AI ^[2]. In the EU's ordinary legislative procedure, a Council position is a significant step: it turns a set of national views into a single negotiating stance, which the Council can then take into talks with the Parliament. Before that point, the member states are still comparing reactions; afterwards, there is a text they have collectively committed to defending.

The Parliament's side

The European Parliament's rapporteurs on the file were Brando Benifei and Dragoş Tudorache ^[1]. A rapporteur is the member who leads a file for a committee: they study the proposal, gather views from other members and outside parties, and draft the report that carries the Parliament's amendments and arguments. Having named rapporteurs signals that the Parliament has taken the file into its own work rather than leaving it with the Council.

Why it matters

Seen from the outside, the most important thing about the April 2021 proposal is that a proposal is not yet a law. It created a shared text that the Council and the Parliament could argue over, and the arguments were about concrete things: what benefits are expected from AI, what risks to privacy, safety and security come with it, and which institution's version of the balance would prevail. The sequence — a Commission proposal, a Council discussion, a formal Council position, named parliamentary rapporteurs — is the ordinary machinery of EU lawmaking, and it is what turned a broad intention to regulate AI into a specific document with dates, authors and amendments attached to it.

For anyone following AI policy, the file is also a record of how the debate was framed at the start. The words used in the Council's discussion — benefits, privacy, safety, security — are the terms in which the later negotiation was conducted ^[2].