Is the EU Coming to Agreement on AI Regulation?

In this update bulletin, Enzai explores the latest in EU AI governance efforts.

‍

The EU is one of the fastest-moving jurisdictions when it comes to the vital area of evolving regulations for the usage of AI. The bloc’s AI Act is in the final stages of its development and is expected to be finalised before the end of 2023.

‍

That piece of legislation, which is widely expected to set the global standard for AI regulation and governance, is currently in the ‘trilogue phase’. Here its exact form is debated between the EU Parliament, which passed its version of the Act back in June, the Council and the Commission.

‍

These ongoing debates, and the resulting lack of clarity, give rise to a question over whether there is, at this point, a clear and united EU approach to the pressing need for AI regulation. 

‍

Disappointment in Granada

The Granada summit, which took place at the beginning of October, presented an opportunity for EU leaders to develop and agree a unified line on the future of AI governance in the bloc, alongside the ongoing trilogue negotiations. 

‍

An informal meeting of the European Council, parallel to that of the wider European Political Community, was foreshadowed as an opportunity for leaders to discuss issues concerning ‘Strategic Autonomy’ - of which AI is a very important component - at the highest levels. It was widely hoped that a united position on AI might emerge from the sidelines of the summit.

‍

Although discussions on AI did play a central role, especially a roundtable discussion attended by major leaders, no joint view on the future of regulation and governance emerged among bloc leaders. The roundtable was jointly chaired by Ulf Kristersson, Sweden’s Prime Minister Edi Rama, Prime Minister of Albania. 

‍

The event, which also hosted non-EU leaders such as the UK’s Prime Minister, Rishi Sunak, was expected to crystallise disparate approaches to AI governance amongst the EU and its close neighbours yet ended only with platitudes such as stressing the importance of intergovernmental collaboration. 

‍

An Italian Approach

While little unity on the future of European AI governance emerged from Granada, some countries have begun pushing their own views of how the bloc’s regulatory efforts should evolve. Italy’s Prime Minister, Georgia Meloni, who also participated in the summit roundtable, has placed herself at the forefront of the governance debate. 

‍

At an AI conference in Italy, after the Granada summit, Meloni declared that it would be an ‘unforgivable mistake’ to not apply rules to the future use of AI. In order to do so, the Prime Minister declared that she would use Italy’s Presidency of the G-7 next year to discuss AI regulation and governance at the ‘highest level’. 

‍

Meloni also set out her views on how this ought to be accomplished. She stressed the importance of centering people, rather than AI, in regulation, which she has conceptualised as ‘algorethics’. While she initially stressed the possible application of such a framework to the labour market, Meloni has yet to set out her ideas in further detail.

‍

The Netherlands Team Up with UNESCO

In contrast to Italy’s more unilateral approach, the Netherlands has opted for greater cooperation in an attempt to further define the EU’s approach to AI governance. The Dutch Authority for Digital Infrastructure has announced a collaboration with UNESCO to launch an AI supervision project. This is further supported by the European Commission.

‍

The project is aimed towards developing the tools and processes which will be required by the EU’s national agencies to monitor the compliance of AI with the requirements of both the EU AI Act and UNESCO’s Recommendation on the Ethics of AI.

‍

It is hoped that this joint project will not just aid the EU in enforcing its AI regulations once they come online, but that it will also strengthen the hoped for ‘Brussels effect’ and help the EU in setting the direction for AI governance globally.

‍

Spain Pitches the Implementation of Obligations

Earlier this week, in its capacity as current holder of the revolving Presidency of the European Council, Spain waded into the EU AI Act debate by emphasising the role of regulatory obligations on AI governance.

‍

The country, whose Prime Minister, Pedro Sanchez, has been at the heart of recent AI regulatory efforts in the bloc, has put forward a draft series of obligations for general purpose AI and foundation models, which would form part of the EU AI Act.

‍

These obligations include the need for affected models and systems to be transparent to regulators, comply with copyright laws and publish a summary of training content, among other responsibilities. Spain’s proposals will be further discussed in continuing trilogue meetings over the coming weeks and months.

‍

Looking Forward

As our update bulletin has made clear, the EU is still searching for unity in its approach to AI regulation and governance. The ongoing trilogue process, which is aiming to come to agreement by the end of the year, should give much more clarity on this issue. 

‍

Time is of the absolute essence for the EU AI Act. Serious efforts are being made to close out the trilogue during Spain’s ongoing Council Presidency (which runs until December 31st). If this is not achieved, the Act’s drafters face an even harder deadline in the form of the upcoming European Parliamentary elections, which are scheduled for early June 2024. If this deadline is unmet, passage of the Act will be severely disrupted.

‍

It is highly doubtful, however, that eventual agreement on the contours of the EU AI Act will close this debate for good. Individual member states will continue to want to have input into a crucial, and growing, area of regulation. Moreover, in shifting the focus already to Italy’s 2024 G-7 Presidency, Prime Minister Meloni has indicated that the future of AI governance will continue to be a topic on global agendas.