Final month, an knowledgeable roundtable passed off convened by Freshfields’ EU Regulatory & Public Affairs workforce primarily based in Brussels, along with international firms from totally different sectors and one of many European Parliament’s lead negotiators on the primary ever regulatory framework on Synthetic Intelligence. Along with German Member of the European Parliament (MEP) Axel Voss, individuals mentioned the state of play of the negotiations on the AI Act, the main focus areas of the textual content and what might be anticipated over the subsequent months contemplating that the Spanish Presidency of the Council has made reaching a last settlement earlier than the top of this yr a prime precedence.
The roundtable in Brussels was organized by Freshfields, hosted and led by Christoph Werkmeister (Associate at Freshfields, World Co-Head of the Knowledge and Expertise Apply) and Natalie Pettinger Kearney (Head of the EU Regulatory & Public Affairs at Freshfields).
The AI Act proposal units up a legislative framework primarily based on threat: the upper the perceived threat, the stricter the rule. To be extra exact, the AI Act proposes totally different ranges of threat – unacceptable threat, excessive threat, restricted threat and minimal threat – which embody totally different obligations for AI suppliers, customers and producers (see Freshfields shopper briefing on the preliminary proposal right here).
Total the AI Act proposal was unveiled by the Fee in April 2021 with the intention of addressing perceived dangers linked to AI and with the intention of turning into a world commonplace setter on this discipline. The truth is, because the EU is the primary jurisdiction to control AI, this proposal may function a blueprint for different regulatory authorities world wide. Following greater than two years of intense negotiations between the co-legislators, the AI Act is at present underneath the ultimate section of the EU legislative course of; the interinstitutional negotiations (trilogues). Whereas the finalisation of the AI Act earlier than the top of the yr is a prime precedence for the Spanish Presidency of the Council, we must always solely count on the AI Act to begin making use of as of 2026.
From the most recent co-legislators’ positions and the present dialogue in trilogues, it’s clear that there are nonetheless many excellent points to be agreed, various which had been mentioned with Freshfields and the corporate representatives who attended to the roundtable. Beneath we set out the important thing points that had been highlighted throughout the roundtable as being probably the most vital.
Defining AI
One of many key dialogue subjects was the definition of AI programs given there was some concern about the truth that the definition proposed by the Fee was very broad. How AI is outlined on this proposal is extraordinarily necessary as a result of it would decide which actions and sectors can be captured underneath the AI Act. Total there was settlement about the truth that the definition of AI needs to be aligned with worldwide requirements reminiscent of these from the OECD or the NIST. It’s as much as the Spanish Presidency to design a compromise proposal that finds the correct stability for EU policymakers, residents and companies whereas allowing for that the OECD can also be revising its personal definition of AI.
Excessive-risk categorisation and record of prohibited practices
A part of the dialogue additionally targeted on the prohibited practices and the record of high-risk categorisation given there are specific disagreements amongst MEPs, but in addition between the ultimate positions of the Parliament and the Council. It was agreed that the potential ban on biometric identification can be a key scorching subject for the trilogues, however the truth that some MEPs usually are not snug with the Parliament’s place may weaken their bargaining energy when agreeing on a compromise with the Council which appears to have a extra stable place on this particular facet. In the course of the dialogue, it was additionally flagged that a number of further necessities had been created for the system itself and for the deployers which may create administrative burden for companies.
Interlinkages with different items of laws
The thought of an current battle between GDPR and the AI Act on the subject of sandboxes was additionally floated throughout the dialogue. In response to a number of individuals, for an AI system to not be biased and discriminatory, it might must be educated with a number of private knowledge, which can result in conflicts with the GDPR which stipulates an information minimisation and a function limitation precept (i.e. private knowledge collected for one function shall typically not be used for different functions, except sure necessities underneath the GDPR are met). Whereas the Parliament has included a provision which explicitly permits the usage of sure private knowledge for various functions, some stakeholders had issues about how this might work in follow (see Freshfields newest blogpost on the Parliament’s place right here).
As a comply with as much as the widespread use of generative AI programs, the Parliament has added an extra layer of regulation underneath the AI Act by making a distinction betweenfoundation fashions and Generative AI programs. This distinction has additionally raised frictions with different items of laws, particularly within the discipline of copyright. In opposition to this background, it was identified that the Parliament has launched further transparency necessities for these programs, in addition to safeguards towards producing unlawful content material.
As anticipated, the dialogue was very open and energetic. Certainly, German MEP Axel Voss was eager to listen to from firms throughout sectors (e.g. monetary, automotive, tech) what the principle challenges for the totally different sectors represented within the room are, and shared views on what Europe ought to do to be on the forefront of digital regulation while preserving innovation. If you want to know extra about what was mentioned and be thought-about for the subsequent roundtable, please don’t hesitate to achieve out to the Freshfields EU Regulatory & Public Affairs workforce.
