ARTICLE 19 joins 40 partner human rights and digital rights organisations and advocacy groups in writing to the European Parliament regarding forthcoming negotiations on the AI Omnibus, a digital package that addresses the implementation of AI rules.
We argue that the AI Omnibus weakens the AI Act, failing to protect people from high-risk AI systems including biometric identification.
We urge the European Parliament and European Council to protect core fundamental rights and safety guarantees and call for the full re-establishment of the transparency requirements under the AI Act.
The letter follows.
Dear Members of the European Parliament,
Dear Cypriot Presidency of the Council of the EU,
Dear representatives of the Member States in the Council of the EU,
Trilogue negotiations on the AI Omnibus proposal are imminent; we, the undersigned experts and organisations committed to protecting fundamental rights and promoting consumer protection, wish to express our deep and persistent concerns both with regard to its substance and legislative procedure.
The AI Omnibus is procedurally deeply flawed. The proposal by the European Commission goes far beyond the purported mandate of ‘technical changes’. This is particularly concerning as the AI Act is a novel law, and the European Commission has not followed basic democratic procedures and failed to follow its own Better Regulation commitment. Specifically, it did not produce any impact assessment, follow a proper review process or conduct an adequate public consultation process for its proposal.
Going far beyond its mandate, the AI Omnibus effectively weakens the AI Act and leaves people in the EU without adequate and timely protection from high-risk AI systems, such as biometric identification or AI use in schools. Over 133 civil society organisations and leading experts previously urged the Commission not to reopen the AI Act following the announcement of the AI Omnibus. They cited grave concerns over lower protections for people and an inadequate process, which, regrettably, have now materialised. The weakening of the AI Act’s safeguards is evident in the positions of all three institutions: the Commission, the Council, and the European Parliament, which – to different extents – introduce sweeping changes that compromise core safeguards and fundamentally alter the structure of the law.
We urge the European Parliament and Council to reject revisiting the protection of core fundamental rights and safety guarantees.
Some of the most concerning changes include the Commission’s proposed deletion of the Article 49(2) transparency safeguard for high-risk AI systems. Therefore, we welcome the position of the European Parliament and the Council on this matter. They reintroduce this vital requirement for registration of high-risk AI systems in an EU-wide database, including providers who self-assess their system as not high risk. However, the positions by both the European Parliament and the Council still exclude the entry of readily-available information from the database, such as the list of EU countries in which the systems are placed on the market and the reason for self-exemption. These exceptions complicate, rather than simplify, effective oversight. Due to these concerns and the remaining threat of the Commission’s proposal to delete Article 49(2) safeguards, we call for the full re-establishment of the transparency requirements under the AI Act.
Further, the governance and oversight powers of fundamental rights bodies are seriously weakened by the current Omnibus positions of all three institutions. These bodies play a critical role in monitoring the use of very sensitive high-risk AI systems, including those deployed by law enforcement agencies. The proposed changes restrict the fundamental rights bodies’ capacity to directly request essential documentation from companies and public institutions. This risks undermining both their independence and their ability to carry out their responsibilities effectively.
Moreover, the changes to Annex I proposed by the European Parliament are alarming. These proposed changes erode the AI Act’s horizontal coherence by altering the structure of the law, for instance, by introducing thirteen delegated acts. Importantly, this complicates, rather than simplifies, the AI Act. Critically, these amendments to Annex I also weaken protections for people in the EU, particularly in sensitive areas such as medical AI. Specifically, the proposed changes undermine the baseline of fundamental rights protections for high-risk medical AI systems introduced by the AI Act.
We therefore call on the European Commission, the European Parliament, and the Council to reject the AI Omnibus on grounds of procedure and substance, and thereby contribute towards ensuring a democratic process, as well as safeguarding fundamental rights protections.
We remain at your full disposal for constructive dialogue and engagement, and we call on the Commission, the European Parliament, and the Council to honour their responsibility in upholding and safeguarding the integrity of the AI Act and its implementation without delay.
Sincerely,
Access Now
AI Accountability Lab
AI Forensics
AlgorithmWatch
Alternatif Bilisim
Amnesty International
Amnesty International Italia
ARTICLE 19
Asociația pentru Tehnologie și Internet (ApTI)
Associazione Comunicazione Pubblica
Bits of Freedom
Border Violence Monitoring Network
Danes je nov dan
Digital Rights Ireland
Digitalcourage e.V.
Digitale Gesellschaft
Equinox Initiative for Racial Justice
European Center for Not-for-Profit Law
European Digital Rights (EDRi)
European Network Against Racism
ESWA – European Sex Workers Rights Alliance
Glitch
Hermes Center
Homo Digitalis
info.nodes
Initiative für Netzfreiheit
IT-Pol Denmark
Italiani Senza Cittadinanza
K-Monitor Association
La Quadrature du Net
Lafede – justícia global
Liberties
Panoptykon Foundation
Period Think Tank
Politiscope
Privacy Network
Rete Diritti Umani Digitali
Sieć Obywatelska Watchdog Polska
Sloweb
StraLi Strategic Litigations
The Good Lobby Italia