This document, still under discussion, seeks to amend and adjust Regulation (EU) 2024/1689, commonly known as the AI Act, in order to simplify its implementation and address a number of early operational shortcomings.
This initiative is particularly revealing: although the AI Act has only recently entered its progressive phase of application, EU institutions are already considering early corrective measures, aimed at striking a workable balance between compliance, fundamental rights protection, and economic feasibility.
This is neither binding legislation nor academic commentary.
It constitutes a formal parliamentary opinion, intended to shape a forthcoming amending regulation within the EU legislative process.
Its objectives are twofold:
A major development is the explicit recognition of AI agents within the scope of the AI Act.
The Committee considers that current definitions may not fully capture systems capable not only of producing outputs, but also of acting autonomously within physical or digital environments.
The proposal therefore expands the definition to include systems producing:
decisions or actions influencing a physical or virtual environment.
The draft opinion proposes a significant amendment to Article 4, arguing that a broad AI literacy obligation is disproportionate and may become overly formalistic.
The Committee suggests replacing it with a more incentive-based approach relying on guidance and best practices.
The draft provides clarifications regarding the processing of sensitive data to mitigate bias.
It proposes extending the legal basis beyond providers of high-risk systems, subject to strict safeguards and alignment with Article 9 GDPR.
The Committee recommends explicitly prohibiting AI systems enabling the generation or manipulation of sexualised content facilitating non-consensual dissemination.
The proposal also seeks to reduce administrative burdens, including certain registration requirements and overly prescriptive implementing acts.
The role of regulatory sandboxes is reinforced, with improved EU-level coordination and SME accessibility.
The Committee considers postponing certain high-risk obligations until December 2027 or even August 2028, due to the absence of harmonised standards.
This draft opinion confirms that the AI Act is an evolving regulatory framework, already undergoing practical consolidation.
A case to follow closely, as these amendments may foreshadow the first substantial recalibration of EU artificial intelligence law.
