AI Act: where do we stand?
European Audiovisual Observatory
On 21 April 2021, the European Commission (EC) presented a proposal for a Regulation laying down harmonised rules on artificial intelligence and amending certain Union legislative acts (Artificial Intelligence Act - ‘AI Act’) (see IRIS 2021-6:1/25 presenting the Commission’s proposal).
Since then, the Council of the EU and the European Parliament (EP) Committee on Internal Market and Consumer Protection (IMCO) and on Civil Liberties, Justice and Home Affairs (LIBE) have started their work. Although the European Commission published its proposal about two years ago, substantial digital evolutions (e.g. OpenAI and its diverse services) have erupted in the course of the last months, providing the institutions with plenty generative AI content to reflect on. In Ferbuary 2023, the institutions committed themselves to ensure substantial progress on the AI file in their joint Declaration 2023-24, and agreed on its high level of priority.
The text that the institutions will negotiate in the future follows a risk-based approach: it will establish obligations for providers and users depending on the level of risk of AI - as identified by the future Regulation.
State of the process:
The Council of the European Union adopted its General Approach in December 2022 and the European Parliament is likely to adopt its negotiating mandate with a view to adopt the AI Act during the June plenary session. Once the latter has adopted its negotiating mandate, the institutions will enter negotiations (the so-called ‘trilogues’) with a view to reaching an agreement on the EC’s proposal.
IMCO and LIBE were appointed as the responsible joint committee in December 2021, with Member of the European Union Parliament (MEP) Brando Benifei (IMCO, Italian, S&D group) and MEP Dragoş Tudorache (LIBE, Romanian, Renew group) as rapporteurs of the text. Together, they prepared a draft negotiating mandate, which IMCO and LIBE MEPs agreed on (vote in committee), on 11th May 2023.
The European Parliament aims to adopt a horizontal text that is present-efficient and future-proof. In the current compromise text, the MEPs agreed on the types of AI and their corresponding obligations. Among the list, we can count: forbidden AI services that bear unacceptable risks (e.g. real-time remote biometric identification in publicly accessible places), high-risk AI with strict obligations (e.g. recommendation systems used by social media platforms categorised as VLOPs under the DSA), and general-purpose AI (e.g generative AI services such as ChatGPT). Generative AI shall abide by transparency rules that may impact the audiovisual world in the future: MEPs want the systems to disclose content generated by AI, and be built in a way that prevents from generating illegal content. Besides, generative AI services shall publish summaries of copyrighted data used for training the programmes.
Since the text is very much horizontal, it could require institutions to set up vertical rules, which could be more specific and adapted to relevant-related topics (such as cultural and creative industries).
It is worth noting that no more than five other committees presented opinions on the text during last summer 2022, among which the Committee on Culture and Education (CULT). CULT suggested amendments obliging AI systems to be transparent when they recommend, disseminate and order news or creative and cultural content (see amendment 55).
Each of these opinions will be taken into account in the final version of the EP’s negotiating mandate.
Council of the EU:
The Council of the European Union adopted its General Approach on the AI Act in early December 2022. One can note that member states of the EU would like to prohibit the use of AI for social scoring by private actors. Furthermore, although prohibited, real-time remote biometric identification in publicly accessible spaces could exceptionally be allowed for law enforcement authorities.
All MEPs are invited to debate the joint committee’s draft during the June plenary session (12-15 June 2023), and consequently vote on the draft negotiating mandate (endorse or reject it). Once adopted in plenary session, the report and the opinions will together represent the EP's position ('negotiating mandate'). Interinstitutional negotiations should follow, using the Commission's Proposal read in the light of the Council's General Approach and the EP's negotiating mandate.
- Proposal for a Regulation of the European Parliament and of the Councillaying down harmonised rules on artificial intelligence and amending certain Union legislative acts
- Council of the EU: Proposal for a Regulation on AI (General Approach)
- Joint Declaration 2023-24
- Opinion of CULT for IMCO and LIBE on the proposal for a Regulation on AI
- Committee report tabled for June plenary
This article has been published in IRIS Legal Observations of the European Audiovisual Observatory.