Timeline and developments AI Act

A more general overview of laws and regulations can be found on this page.

Adopted text

The adopted text of the AI Act can be found here. This is the final text adopted by the European Parliament, as of the 13th of March 2024.

Recent votes

On the 13th of March the Members of the European Parliament voted to approve the result of the negotiations with the Member States on the AI Act. It was endorsed by the MEPs with 523 votes in favour, 46 against and 49 abstentions. All that is left is the linguist check and final Council endorsement. 

On the 13th of February the Internal Market and Civil Liberties Committees voted to approve it as well. The most recent version of the AI Act (2 February 2024) can be found here. This is the provisional agreement resulting from interinstitutional negotiations. 

Compromise Text

The negotiations have concluded and the Parliament and Council have come to a compromise text with view on agreement. The Council has published their analysis of the text on the 26th of January.

ARTICLES ON THE DRAFT AI REGULATION

• AI Act flowcharts
• Pre-final AI Act: updates
• High-risk applications of AI systems
• Requirements for high-risk AI systems
• Prohibited applications of AI systems
• AI according to the European Commission

For updates on the AI Regulation, subscribe to our AI newsletter here.

CRITICAL REACTIONS TO THE PROPOSAL

ITechLaw has published a green paper on the AI Regulation proposal. The paper makes a number of criticisms of the proposal.

First, the authors stress that the proposal is not a standalone law. The proposal leans on the GDPR and also any future laws on AI liability. This creates a lack of certainty for parties who have to start complying with the proposal's requirements.

Second, the proposal takes a compliance approach. An AI system has to be audited in a certain way and must meet a list of requirements. This may result in a particular system not being able to meet any of the requirements and thus not being approved, even though by practical standards it is working well.

Thirdly, it is not yet clear how citizens can claim their rights and, for example, get an explanation of a decision. This is currently already possible through the GDPR, but the AI Act does not seem to add any new possibilities.

Finally, the authors of the paper conclude that the structure of the proposal is too complex. The various regulators and other interested parties who have to make decisions together and the requirements for high-risk systems could make AI systems too expensive for SMEs, especially startups. Many of the proposal's requirements also rest on individual developers, which may lead to a difference in quality of implementation.

NEGOTIATIONS IN THE EUROPEAN PARLIAMENT

While negotiations are taking place in the Council by representatives of member state governments, the Parliament is also going to consider the Commission's proposal. The IMCO Committee (Internal Market and Consumer Protection) and the LIBE Committee (Fundamental Freedoms) will lead the negotiations. These expect to have the first draft report ready by April 2022. After that, there will be three weeks to propose amendments. The vote on the proposal is scheduled for 29 September 2022.

The draft opinions of the following committees are now open for comments:

• Committee on the Environment, Public Health and Food Safety (10-12-2021)
• Committee on Culture and Education (21-2-2022)
• Committee on Legal Affairs (2-3-2022)
• Committee on Industry, Research and Energy (3-3-2022)

The Committee on Legal Affairs published a second version of its Opinion on the AI Act.

NEGOTIATIONS IN THE COUNCIL OF THE EUROPEAN UNION

The Council of the European Union began negotiating the European Commission's proposal in July 2021. The proposal is being discussed in pieces. In the first round, articles 1-7 and annexes I-III were discussed under the leadership of the Slovenian presidency. The second round began in January 2022 under the French presidency. In it, articles 8-15 and annex IV were discussed. In February 2022, the Council then discussed Articles 16-29 and also Articles 40-52. In March, it was the turn of the Regulatory Sandboxes, article numbers 53 to 55. Articles 30-39 and 59-62 were then dealt with in April. The French started the final sprint in May, with Articles 70 to 85, Articles 56-58 and 63-69 and also Articles 3 and 4 (not yet available online).

The Council has combined all its changes and additions in a consolidated version. During the presidency of the Czechs, the definition of High Risk systems is being tinkered with (link to be added later). High Risk systems must be on one of the lists of High Risk applications, meet the functional requirement and then also have a significant impact on the decision they affect. This is a narrowing of the scope compared to previous proposals.

FIRST CHAMBER CONSIDERATION

The Senate submitted questions on the Regulation to the Cabinet and to the European Commission during the end of 2021. To these, the Cabinet replied on 14 December 2021. The European Commission replied on 5 February 2022, explaining some of the trade-offs involved in writing the Regulation.

TREATMENT CABINET AND SECOND CHAMBER

On 10 September 2021, the House of Representatives' Standing Committee on Digital Affairs submitted its questions to the cabinet on the Regulation on Artificial Intelligence and the Communication Promoting a European Approach to Artificial Intelligence.

On 31 May 2021, the foreign minister presented the cabinet's views on these two EU proposals to the House of Representatives. This summarises the proposals and explains the Cabinet's questions. (Vision of the European approach, Vision of the AI regulation)

EU

April 2021 it was finally the moment that the European Commission presented a policy review, Fostering a European approach for AI, the proposal for the AI regulation and an updated plan for coordinating the approach to AI.

Through 2020, these documents were subjected to a process of public consultation and discussed and reviewed many times.

Around February 2020, the EU already became more concrete with the publication of the White Paper on AI: a European approach to excellence and trust and then in July with the Assessment List on Trustworthy AI by the High Level Expert Group on AI.

In April 2019, the process towards AI regulation began with the publication of the Ethical Guidelines for Trustworthy AI by the EU's High Level Expert Group on AI.