login
login
Image header Agence Europe
Europe Daily Bulletin No. 13802
Contents Publication in full By article 11 / 32
SECTORAL POLICIES / Digital

Revision of AI Act - a first parliamentary report with no real brake on European Commission’s ambitions

The European Parliament has been more incisive in the past. The initial version of the report on the review of the AI Act, which was expected to be the first political response by elected representatives to the Commission’s drive for simplification and its digital package (see EUROPE 13755/3), has so far left a mixed impression.

The document, published by Agence Europe, makes minor amendments to the initial proposal, which had already been severely restricted by the Member States (see EUROPE 13794/8). One potential consequence of the very tight timetable: the rapporteurs had less than three weeks to prepare their version.

More protection for personal data. Far from matching the level of rewriting in the EU Council’s version, the amendments in the parliamentary report are short and focus on two strategic points: a fixed deadline for the entry into force of the obligations for high-risk AI systems, and a restriction on the flexibilities for the use of personal data by AI models. 

Like the Member States, Parliament’s rapporteurs are reviving the notion of “strict necessity” in the new Article 4a, which allows suppliers of high-risk AI systems to process special categories of personal data. Its removal by the European Commission was strongly criticised by many players in the sector (see EUROPE 13786/15)

The second paragraph of Article 4a is also amended. This opens up the possibility for other AI models and deployers to also process particular categories of personal data. The text restricts this possibility “on an exceptional basis” only.

Fixed maturities for high-risk models. The parliamentary report, like the first EU Council compromise, removes the possibility for the Commission to trigger the application of these obligations prematurely. MEPs retained the two ‘extended’ deadlines, i.e. December 2027 for the first category of high-risk models, and August 2028 for the second.

In the original text of the AI Act, these obligations were to come into force on 1 August 2026. However, faced with the delay in drawing up many of the technical standards, common specifications and guidelines associated with these obligations, the Commission has agreed to postpone the deadlines.

This postponement is included in the revision of the Regulation and, in order to be effective, must be validated by all the co-legislators before 1 August this year. 

A challenge for a text that represents a technical and political challenge, both for the member countries and for certain members of parliament. The Cyprus Presidency of the Council of the EU, which is responsible for moving the text forward, hopes to organise the first interinstitutional negotiations by April, although it is not certain that this will happen. 

The MEPs on the two parliamentary committees responsible for substantive issues (IMCO and LIBE) have until 12 February to table amendments to the text. 

See the report: https://aeur.eu/f/klx (Original version in French by Isalia Stieffatre)

Contents

EXTERNAL ACTION
SECURITY - DEFENCE
INSTITUTIONAL
SECTORAL POLICIES
ECONOMY - FINANCE - BUSINESS
COURT OF JUSTICE OF THE EU
COUNCIL OF EUROPE
NEWS BRIEFS
Op-Ed