The future of AI in the European Union: a chat with the AI Act’s co-rapporteur Brando Benifei

We recently sat down with MEP Brando Benifei (S&D, Italy), co-rapporteur on the AI Act and member of the parliamentary committee on the Internal Market and Consumer Protection (IMCO) and the Special Committee on Artificial Intelligence in a Digital Age (AIDA).

It was a unique occasion for us to ask our burning questions on the recently adopted AI Act, which marked a significant step forward in regulating AI in the EU. This interview explores how the AI Act aims to strike a balance between fostering innovation in the European AI landscape and mitigating potential risks for consumers, workers, and organisations.

MEP Benifei shared his insights on the consequences of the AI Act on business competitiveness, AI governance in the EU and beyond, and the future of AI regulation in the EU.

 

The AI Act’s effect on European competitiveness

Our readers have been following the legislative developments of the Artificial Intelligence (AI) Act very closely. Could you explain why this regulation matters in the race for a more competitive Europe? There have been some speculations, particularly in Italian newspapers, that the AI Act is going to put additional burdens on the economic activities and the industry here in Europe. How would you address such worries?

 

MEP Brando Benifei:

First of all, we have to say that the EU has been lagging for some time behind the US and China on AI development. The AI Act has nothing to do with that. This delay is due to the level of investment, research, and access to data. At the moment, the EU cannot compete with the US Inflation Reduction Act (IRA). The issue now is to understand the impact of the AI Act: would it make the situation better or worse for EU competition?

To this I would reply: it depends on what we do next. The EU has to show unity in addressing pending issues such as the need for supercomputing power, the completion of the capital markets union, promoting investment and resources, and the need for an industrial policy strategy that boosts AI development. At the same time, the AI Act should spark commitments here in Europe to protect consumers, workers, SMEs, democratic institutions, and build public trust in AI technology. This is the objective of the Regulation: building more trust for more uptake of AI. And by building trust, we could also trigger more political will to invest in AI competitiveness at European level.  

On another note, it is good to remind that the AI Act applies to businesses anywhere in the world that make use of the European market, and not just EU businesses. As it happened before with the GDPR (i.e. General Data Protection Regulation), I feel that there is no way that the AI Act will dissuade companies from third countries to invest in the European market.

The AI Act’s impact on your business’ daily activities

Evidently, the AI Act is an important file, but could you explain to our readers its impact on day-to-day business and life?

 

MEP Brando Benifei:

As previously mentioned, the AI Act should spark commitments here in Europe to protect not only consumers and workers, but also SMEs. The AI Act aims to support through its provisions even new businesses, for example, when we deal with the issue of sandboxes. As such, the AI Act supports the possibility for businesses to develop new ideas and entering into the European market.

I think we also need to underline that provisions such as the conformity assessment that is required for the most sensitive use cases of AI, is in most cases a self-assessment that has to be done by the businesses. This has been done to not put too many bureaucratic burdens on businesses.

Instead, we asked them to act responsibly, and do their checks on the requirements that we have put for the most sensitive use cases of AI that are deemed more dangerous. And also, I need to underline one more time that these rules apply to businesses anywhere in the world that make use of the European market, and not just EU businesses.

Global AI governance

Digital ministers from G7 countries under the Italian presidency committed to finding ways to monitor compliance with the so-called Hiroshima Process voluntary code for artificial intelligence. The code includes pledges for ethical AI development, transparency, and risk mitigation, including watermarking AI-generated media. Italy seems to want to lead on the subject of AI in Europe. News outlets also reported Italy’s plan to set up an investment fund of 1 billion euros to promote AI projects. Can you guide us through the Italian government’s ambitions on AI and explain why Italy wants to lead Europe and the G7 with regards to AI?

 

MEP Brando Benifei:

My sincere hope is that the G7’s work on AI will be meaningful, and not just a space for empty promises. Indeed, the Italian government seems to be positioning itself in a leadership role on AI in the context of the G7 presidency.

However, in terms of concrete steps that the Italian government has taken so far on the issue, we have been seeing only promises for the last year or so: promises of fiscal incentives, public-private partnerships, national rules to complement the AI Act. But nothing has come out yet.

I honestly do not believe the Italian government will come out with additional rules on AI at national level, but I think the proposal of fiscal incentives could be supported as a good provision. Now that the AI Act is in place, I hope that European national governments, including the Italian one, and the European Union will do their part in supporting investment and public-private partnerships.

At the same time, we see that on AI there is a need for international governance that goes beyond just what we do in the EU. This is due to the fact that, at the moment, we face challenges of a global nature (transparency, data sharing, data security) which we need to address in different structures with different focuses depending on their strengths.

We talk about efforts on AI threats in multilateral fora such as the Organisation for Economic Co-operation and Development (OECD), the United Nations (UN), the Bletchley Park conference, but also the work done so far under the G7 for the Hiroshima process, which I believe will spill over to any G20 discussion and work on AI.

It’s important to have these initiatives focusing on different issues linked to the use of AI, which will then need to align to secure more comprehensive international cooperation framework and rules for AI that cannot be dealt with via domestic legislation alone.

European AI Governance

The governance of AI in Europe is already shaping to be rather complex (AI Office, AI Board, Scientific Panel, Advisory Forum, partnerships and fora with stakeholders, AI Pact and more). What would your advice be to companies using AI or aiming to start using such technology in a safe and (AI Act-)compliant way? Who should they call for information and clarifications on the law?

 

MEP Brando Benifei:

I would like to provide three venues that businesses may wish to take into consideration. First of all, stakeholders should get involved as soon as possible in the AI Pact* that is now being rolled out. This initiative will support both businesses and institutions that develop or use AI in learning how to start and anticipate compliance to the AI Act requirements.

For the AI Pact, the Commission will start a tour in November in all EU member states to meet with business organisations and national ministries. We also want both member states and the Union to support this scheme with incentives, which in turn can bring higher compliance rates.

Secondly, businesses may call on the Commission to present their expert support in implementing the AI Act. As a result of the AI Act, the Commission has established an AI Office**, which is now hiring technical and expert staff.

Thirdly, in the current mandate, the European Parliament has been facilitating contact between organisations and the Commission on what to do next with regards to the AI Act and compliance to it. We expect that the Parliament will continue to supervise the implementation of the AI Act also in the new mandate.

* Note from Publyon: the AI Pact enables opportunities for businesses to help with (early) compliance with the AI Act. Should you be interested, Publyon can support your organisation in getting involved in the Pact.

** Note from Publyon: established in February 2023, the AI Office will, among other tasks, oversee the implementation of the AI Act, support and coordinate secondary legislation, and oversee the AI Pact.

What to expect for AI after the European elections

The European elections will take place in June, the first step for a new mandate with new political priorities. AI is bound to remain a topical issue for the EU institutions. Can we expect more policy initiatives on AI, e.g. on copyright, deepfakes, skills, IP protection, R&I, etc.?

 

MEP Brando Benifei:

In 2022, the former European Parliament’s Special Committee on Artificial Intelligence in a Digital Age (AIDA) formulated a report with recommendations to the European Commission regarding AI. In the report, the AIDA Committee recommended the Commission to come up with a proposal for a directive on AI at the workplace, which could not be dealt with under the AI Act due to competence matters.

My expectation is that the European Commission will seriously examine and proceed on this specific policy initiative in the first half of its next mandate, which will also see the implementation and monitoring of existing legislation, including the AI Act.

In addressing the topic of copyright in relation to AI, I personally do not see any new policy initiatives on this topic coming soon, nor the need for them at this stage. The European Union Copyright Directive (EUCD) and the AI Act already provide frameworks for regulation.

However, the devil lies in the details, as seen with requirements such as the AI Office receiving summaries of datasets used in generative AI development. Monitoring the implementation of the AI Act, particularly concerning copyright, will be crucial in the coming years.

If the current regulations prove insufficient, the EU may consider more specific legislation. The existing regulations aim to apply copyright protection clearly to the AI context, emphasizing the need for effective implementation. Should it become evident that these norms are not adequately protecting creators and the creative industry, revisions may be necessary.