Innovation vs Regulation: Europe’s Response to AI and Platforms

As the rapid rise of Generative AI reshapes the digital landscape, European regulators are grappling with new challenges in antitrust enforcement – and considering whether they can utilise existing mechanisms, or need to build anew. ‘One immediate question is whether and how regulators in Europe and elsewhere […] can rely on our past experience of antitrust enforcement in digital markets to tackle the challenges that GenAI, in particular, poses’, says Fayrouze Masmi Dazi, Founding Partner at Dazi Avocats and a specialist in French and EU competition law. Certainly, the Digital Markets Act (DMA) is already in the spotlight in regards to Generative AI, she emphasises: ‘There is no doubt AI is part of the DMA.’ 

In this dynamic environment, European authorities are emerging as leaders. The Autorità Garante della Concorrenza e del Mercato (AGCM) in Italy and the Autorité de la Concurrence in France are both highly active in assessing and pursuing antitrust cases in their home countries and in conjunction with authorities at the European Commission. As Fayrouze notes, ‘these two […] are among the most active authorities in Europe regarding antitrust enforcement in the digital economy.’ As such, they are uniquely well-positioned to comment on the current state of Generative AI regulation and where it may be headed next.

Revisiting old issues, confronting new ones

‘Artificial intelligence seems to raise not only new issues, but also very traditional concerns,’ says Andrea Pezzoli, Vice Secretary General at the AGCM. ‘It’s a very fashionable issue for antitrust agencies’, he continues, but he also thinks there are many similarities to previous regulatory cases: ‘At the end of day, we are dealing with typical vertical issues […] along the value chain.’

The AGCM has had plenty of experience in regulating digital markets in recent years. One case involving Google and Android, related to electric mobility and services for recharging electric cars, is currently under review by the Italian Court of Justice. Another significant case concerns Amazon’s Fulfilment By Amazon (FBA) service and focuses on tying and visibility issues, where sellers who do not use Amazon’s logistics services face reduced visibility on the platform. Self-referencing and discrimination in privacy policies are features of yet another ongoing case, with involvement from Italian, French and German authorities. This latter case particularly demonstrates how privacy may be used as a tool to potentially exclude competition. 

In Andrea’s opinion, one lesson from this recent work on digital markets that regulators can apply to Generative AI is collaboration between various authorities. ‘Sometimes the languages are not always consistent, so it’s not so easy to cooperate, but it is a must. We have to do it if we want to cope with digital issues or the issues that could be raised by artificial intelligence’, he says firmly. However, he notes that AI also presents different challenges from those in the digital platform world, particularly around market dominance and the potential for monopolies. 

With all this in mind, Andrea asks whether it’s better to implement regulations quickly or take a more cautious approach. ‘We still have to decide if timely regulation is better than more conscious regulation, which may arrive too late,’ he says, while also acknowledging concerns that reactionary regulation could stifle innovation at key stages of Generative AI’s development.

Keeping pace with AI innovations

Meanwhile, the French competition authority recently conducted a detailed study on the AI market. In contrast with an opinion last year that they prepared on cloud services, which took a year and a half to compile, this opinion was published within four months. ‘We see that there is a need to be faster and maybe not tackle all the issues’, explains Elodie Vandenhende, Deputy Head, Digital Economy, at the Autorité de la Concurrence. Instead, Elodie’s team focussed on analysing key issues further upstream.

‘We see that there is a need to be faster and maybe not tackle all the issues’ – Elodie Vandenhende 

The inquiry came to three broad conclusions about the AI market. ‘The first one is you have major digital players that are already active at different levels of the AI value chain, and some of them hold significant positions in several markets’, Elodie explains, citing Google as an example. ‘The second conclusion is that the sector is characterised by significant barriers to entry and expansion, because [it] requires at least four key inputs: chips, cloud infrastructure, data [and] technical skills. These inputs can be rare and or expensive. And finally […], some digital players have a headstart in this sector because of their [competitive] advantages linked to their vertical and conglomerate integration.’ 

According to Elodie, the inquiry also identified anti-competitive concerns in the Generative AI sector, particularly around data access and recruitment of skilled labour, like Microsoft’s hiring spree of key employees from Inflection. Other risks are linked to chip manufacturing and procurement – a competitive advantage for some Generative AI players that could become even more pronounced over time. This observation echoes findings that the French competition authority released in its opinion on cloud service providers last year. ‘We find that some of those risks that we had identified [in regards to cloud service providers] seem to be persisting and even intensifying with respect to Generative AI’, Elodie notes. Strategic partnerships and minority shareholding were also put under the microscope, but the recent opinion includes a clear framework to tackle these issues using competition law. Its ten recommendations underscore the importance of making full use of current regulations and ensuring data access. ‘There are a lot of things that we can do already with the regulations that are in place and our competition tools’, Elodie says.

The need for collaboration is increasingly strong

Indeed, the regulations that are already in place are powerful levers for antitrust regulators. In Andrea’s opinion, the role of competition agencies has been significantly impacted by the DMA, particularly regarding gatekeepers. ‘A lot of potentially abusive conduct has been already included in Article 5 and 6 of the [DMA]’, he says, adding that past anti-competitive behaviours have informed the current framework. However, some measures are still under judicial review, and Andrea thinks that there is still room for antitrust actions on behaviours not covered in the DMA. 

Considering the DMA’s framework, Andrea believes that regulatory bodies across Europe will need to coordinate their efforts. ‘The need for collaboration is increasingly strong’, he says, especially with the increasing complexity of the regulatory landscape.

‘The need for collaboration is increasingly strong’ –   Andrea Pezzoli

Previous cooperation between the AGCM and the European Commission proved successful, such as in a significant case against Amazon, and the Italian authority is currently partnering with French and Swedish ones to address pricing issues in the travel sector. According to Andrea, the AGCM also published draft guidelines during the investigative phase of the DMA, outlining procedures for addressing potential non-compliance, ensuring transparency and confidentiality and managing document access during investigations. These are just a handful of examples of how national authorities can foster and advance effective enforcement of the DMA. ‘If, at the national level, we can provide our support, I think it would be welcome’, Andrea adds.

Similarly, the French authority’s case study on Generative AI built upon existing reviews by other regulatory authorities. ‘[Generative AI] is a sector that is intensely looked at by other authorities, [so] I think you need to try to see where you can add value’, says Elodie. For example, their opinion looked closely at the issue of data access, particularly for newer entrants to the market who may not be able to purchase the volume of data necessary to train a large language model (LLM). They also paid attention to technology that claims to be open source. ‘Open source is a term that is sometimes used very loosely by some actors,’ Elodie thinks. ‘We in the antitrust community know that sometimes players can offer access to an open-source technology and then close access later.’

For her part, Elodie welcomes the shared interest in these issues within both the international and European regulatory communities. ‘Some of these issues that we’re raising also have other authorities thinking about it and contributing to the debate’, she says.

The best approach to Generative AI? A proactive one

The debates are ongoing, but many regulations are already in effect that can impact Generative AI. ‘[We] have a lot of digital regulations that have been passed, mostly over the last two years. Some of them have not really been enforced yet’, notes Elodie. This context also formed an important part of the French authority’s opinion, as they assessed implementation and potential loopholes. One open question based on their findings is if AI markets should be designated as core platform services under the DMA, since such a designation would affect user access to AI models.

Overall, Elodie advocates for an operative approach to regulating Generative AI. ‘The general theme is that we need to be vigilant [and] active, because we have a lot of legal tools at our disposal’, she reiterates. At the same time, she believes it’s important to stay informed about the dynamics of Generative AI to effectively address potential issues.

With G7 meetings on the horizon in October 2024, there seems to be a broader focus on Generative AI among major antitrust agencies. Andrea echoes Elodie’s call to make use of both traditional and new regulatory tools in addressing AI-related challenges, and he notes that the ongoing market investigation into airline pricing algorithms could eventually end up setting a new judicial precedent. But in the meantime, the AGCM is working proactively to better understand the complexities of AI: the agency is strengthening its expertise by hiring data scientists and setting up a technology unit. 

Join the conversation

This panel with Fayrouze Masmi Dazi, Andrea Pezzoli and Elodie Vandenhende was part of the Platform Leaders online event on 11th July 2024 (full list of speakers and agenda), organised by Launchworks & Co.

Platform Leaders brings together hundreds of experts worldwide through its in-person and online conferences, providing a unique forum for entrepreneurs, academics, practitioners and policymakers. Explore the full lineup of speakers and topics, browse articles and videos from past conferences and connect with the community to stay updated on future events and much more.