Locations
In its roadmap for 2024-2025, the Competition Authority has set itself the task of pursuing its commitment to digital issues.
After issuing an opinion on competition in the cloud sector in June 2023, the Authority has just published its first opinion on competition in the generative artificial intelligence sector, a sector that has been at the centre of public and economic debate since the launch, at the end of 2022, of assistants such as ChatGPT, capable of generating texts and images.
In this opinion, the Authority warns companies of the risks associated with the development of this sector, in particular the fact that major digital players such as Google and Microsoft have a significant competitive advantage in this sector. In addition to their financial power, which enables them to enter into numerous partnerships, they also have easy access to the necessary computing power, privileged access to data and the ability to attract the best talent. All these factors together constitute a significant barrier to entry and expansion for competitors, and the advantages enjoyed by the large companies may create risks for competition.
In particular, the Authority is concerned that this competitive advantage could be strengthened by the integration of these large digital players across the entire value chain of generative AI, as well as in separate but related markets in the generative AI sector, which would give them economies of scale and access to users, businesses and consumers. This would therefore lead to a high degree of concentration to their benefit and strengthen their power in separate but related markets such as office productivity software, search engines or online advertising.
In order to mitigate this risk of dominance, the Authority, in accordance with its role under Article L462-4 of the French Commercial Code, makes a number of recommendations aimed at promoting competition in the sector :
- Invite the French AI Commission to examine the possibility of designating companies providing services enabling access to generative AI models as gatekeepers through the prism of cloud computing services as defined in Article 2 of the DMA;
- Encourage the DGCCRF to pay particular attention to the use of cloud computing resources in the field of AI, especially in the context of the implementation of the law of 21 May 2024 on the security and regulation of the digital space;
- Obtain the support of the competent authorities and make use of the tools available. The Authority therefore undertakes to remain vigilant, alongside the DGCCRF, on the behaviour of the main players in the generative AI sector value chain, in order to mobilise, if necessary, all their respective tools to act quickly and effectively;
- Develop public supercomputers as an alternative to cloud providers, and in particular to give academic players access to computing power that is conducive to innovation. The Authority is also in favour of opening them up, under certain conditions, to private operators in return for payment;
- Preserving innovation by striking a balance between fair remuneration for rights holders and access for model developers to the necessary data. The Authority therefore calls on the government to encourage right holders to take into account the economic value of data according to use cases (e.g. by introducing differentiated pricing) and to offer bundled packages that reduce transaction costs in order to ensure the ability of model developers to innovate;
- Achieve greater transparency on minority shareholdings in innovative companies, based on Article 14 of the DMA, which allows designated companies to be asked for information on minority shareholdings in the same sector as the target.
The Authority points out that most of these recommendations do not require legislative initiatives at French or European level.
This opinion is also part of a debate that goes beyond the national framework, as the generative AI sector is also the subject of increased attention at European level, with the European AI Regulation (AI Act), currently being published in the Official Journal of the European Union, which will essentially apply from 2026. This regulation will apply to both public and private actors, inside and outside the EU, as soon as the 'AI system' is placed on the market in the Union or its use has an impact on EU residents.
As a reminder, the AI Act sets out obligations based on a risk-based approach for AI systems. In particular, systems that are deemed to pose an unacceptable risk are prohibited, while 'high risk' systems are subject to specific requirements. The AI Act also imposes transparency obligations on providers of generative AI models and the implementation of measures to ensure compliance with European copyright law.