The use of artificial intelligence in the automation of business processes and improved customer engagement is becoming increasingly common. Consequently, new branches of artificial intelligence such as generative AI, which is a model of artificial intelligence that creates new content rather than manipulating or replacing existing content, are emerging. Being open to innovation and developing very fast, the generative AI sector has been on the radar of competition authorities around the world. It is not yet clear what the competition policy will be for generative AI. Even though competition authorities share similar views regarding the competitive concerns and possible solutions in generative AI sector, it is uncertain whether they will adopt preemptive regulations or monitor the development of market dynamics for a while longer. Under this article, the approach of competition authorities to potential antitrust violations stemming from generative AI and emerging regulatory framework is discussed.
On May 4, 2023, the CMA launched an investigation to determine guiding principles for the future use of artificial intelligence models.1 Similarly, on June 29, 2023, the FTC highlighted antitrust issues associated with generative AI and subsequently opened an investigation under consumer protection laws.2 Similar efforts are being made by the European Union to adapt artificial intelligence systems to the markets and to regulate their use and liability through the Artificial Intelligence Act3 and the Artificial Intelligence Liability Directive4. EU's AI act is considered the world's first comprehensive AI regulation and establishes a broad framework for AI governance across the EU. The AI Act defines an “AI System” as a machine-based system designed to operate with varying levels of autonomy and adaptiveness. The Act classifies AI systems into four risk categories “Unacceptable Risk, High Risk, Limited Risk, and Minimal Risk” based on factors such as potential interference with human rights and employment impacts. The AI Act imposes strict prohibitions on AI systems classified as Unacceptable Risk, while other categories face varying levels of obligations and compliance requirements. Additionally, the Act introduces the concept of General-Purpose AI Technologies, including General-Purpose AI Models (GPAIM) and General-Purpose AI Systems (GPAIS), which are subject to additional obligations due to their broad applicability. A timeline has also been established for companies to adapt to these new regulations, with the AI Act entering into force on 1 August 2024, and full compliance required by 2 August 2027. The steps taken highlight the complex challenges faced by competition authorities due to artificial intelligence tools and provide guidance for other authorities.
Although it has not yet started to be widely used by enterprises, there are some competition risks emerging from Generative AI. CMA's and FTC's publications explore how generative AI interacts with competition law and analyze potential risks. According to the authorities, with the development of generative AI, the following competition law risks are likely to emerge:
- Control Over Essential Inputs: If a limited number of undertakings quickly master important elements of generative AI such as data, capabilities and computational resources they will be able to manipulate or limit competition. Control over these inputs would hinder the competitive environment and the broad benefits of generative AI.
- Data Domination: Large-scale, high-quality data repositories are vital to the development of generative AI. Because productive artificial intelligence works using these data pools. Since well-established technology companies will have a primary competitive advantage in accessing comprehensive user data and tools, there may be obstacles for new undertakings that want to enter the market. FTC also considers the possibility that this risk can be prevented naturally, thanks to the innovation-based structure of the market.
- Monopoly of Talent: The scarcity of employees equipped in the field of generative AI may encourage powerful companies to retain talent, and in this context, anti-competitive behaviors in the labor market may emerge.
- Open-Source Dynamics: Many start-ups that have entered the market with the new and rapid development of Generative AI have shared their resources publicly, free of charge. While this is generally considered as an element that encourages competition, the FTC also reminds that open-source artificial intelligence models are open to abuse and may be subject to "open first, then close" tactics.
- Mergers and Acquisitions: The CMA and the FTC both state that competition authorities should approach mergers and acquisitions especially carefully in markets based on innovation and where many start-ups are formed, such as the generative AI market, as killer-acquisitions may harm the functioning of the market.
More recently, The French competition authority Autorité de la concurrence has issued its opinion on the competitive dynamics of the generative AI sector.5 Autorité's opinion focuses on strategies by major digital companies to consolidate their market power in the generative AI value chain. It also examines potential barriers to entry for smaller undertakings. Similar to CMA and FTC, the Autorité explains in its opinion that generative AI sector has high barriers to entry:
- Firstly, developing foundation models in the generative AI sector requires significant computing power, typically provided by specialized AI chips such as GPUs from Nvidia or AI accelerators from companies like Google. These resources are not only expensive but also in high demand, making them a substantial barrier to entry for new players in the market.
- Secondly, cloud services play a crucial role in the generative AI ecosystem. Accessing the necessary computing power and distributing AI models are primarily facilitated through major cloud platforms like Amazon Web Services (AWS), Google Cloud Platform (GCP), and Microsoft Azure, which are in dominant positions in their markets. According to Autorité, the dependence on these cloud services further raises the entry threshold for emerging companies.
- Thirdly generative AI models need vast amounts of data, which are primarily sourced from public domains. However, there are growing concerns about the future availability of such data and legal issues related to data usage, potentially complicating the data acquisition process for new entrants.
- Fourthly, the development of AI models necessitates advanced data science skills, including expertise in machine learning and deep learning. According to Autorité these skills are rare and highly sought after, which adds another layer of difficulty for new undertakings trying to enter the sector.
- Finally, significant financial investments are required in the generative AI sector. These financial requirements also act as a barrier to entry for startups and smaller companies.
Autorité explains that in addition to these entry barriers, major digital companies have other competitive advantages in the generative AI sector. Since these companies have better access to computing power, data and skilled workforce Autorité believes that they have a significant edge over their competitors. Moreover, their integration across the value chain allows them to leverage economies of scale and scope and to integrate AI tools into their ecosystems, which can further consolidate their dominant positions. It can be seen that competition authorities share similar views regarding the potential entry barriers and the state of major digital companies.
Autorité also examines the possible competition risks that can arise in markets related to generative AI sector. According to Autorité;
- Chip providers could abuse their position by price fixing, production restrictions and unfair contractual conditions.
- Cloud service providers could offer high levels of cloud credits to startups and create technical barriers to migration.
- Restrictive access to data can disadvantage smaller players and restrictive access to skilled workforce could lead to practices like wage fixing and no-poaching agreements.
- Vertical integration could lead to abusive practices such as denying access to necessary inputs for training AI models and tying sales of products to their own AI solutions.
Autorité provides recommendations in its opinion to address these competitive concerns. According to Autorité, to make the regulatory framework for generative AI more effective, the Commission should consider designating companies providing Model-as-a-Service (MaaS) as gatekeepers under the Digital Markets Act (DMA) to prohibit problematic behaviors proactively. Currently, the DMA does not contain specific provisions addressing generative AI, leaving a gap in the current legal framework. This highlights the need for updated regulations to ensure that emerging technologies like generative AI are adequately covered under competition laws. Autorité also recommends that the future AI Office and national authorities should ensure that the AI Act does not hinder small operators or allow large operators to misuse the regulations. According to Autorité, international coordination is also essential to prevent distortions and additional costs for companies.
Generative AI has been on the radar of Portuguese and Hungarian competition authorities as well. Audoridade da Concorrência has published an issue paper about generative AI which focuses on topics like access to data, cloud computing and base models.6 Hungarian Gazdasági Versenyhivatal has started a study in collaboration with market players to analyze importance of data, potential entry barriers and the role of open-source models in generative AI sector.7 Recently the CMA launched an investigation to assess whether the partnership between Amazon and Anthropic, an AI research, security and manufacturing company, might raise any competition concerns.8 While the investigation is still ongoing, with the Phase 1 decision is expected in October according to the CMA's timetable, it highlights that competition authorities are closely scrutinizing the activities and agreements of major digital companies in generative AI sector.
It appears that many countries are examining the interaction between the rapidly growing generative AI sector and competition law. Competition authorities seem to have reached a consensus on issues such as possible entry barriers, the importance of data and the advantages of major digital companies. However, early intervention by competition authorities, especially in innovation-driven markets, can negatively affect the development of the sector, technology, and the benefits consumers derive. Therefore, competition authorities prefer to fully understand the sector by observing it from afar before taking action. Consequently, the OECD warns competition authorities to be vigilant and to refrain from early interventions. While early intervention in a market that could become competitive and innovative on its own may have negative effects, delayed intervention may result in “winner takes all/most” outcomes. In this respect, the OECD likens the development of the generative AI sector to digital markets and advises competition authorities to apply their experience in digital markets to this new sector.9
Nevertheless, based on competition authorities' approach, we can expect the first steps towards establishing regulatory frameworks in many countries in the near future. A proposal for an artificial intelligence law was recently submitted to the Parliament in Turkey.10 The proposal aims to ensure the safe, ethical and fair use of artificial intelligence technologies, to ensure the protection of personal data, to prevent the violation of privacy rights and to establish a regulatory framework for the development and use of artificial intelligence systems. As competition authorities around the world continue to study the complexities and potential risks associated with generative AI sector, they try to strike a balance between fostering innovation and ensuring fair competition.
Footnotes
1. CMA, CMA launches initial review of artificial intelligence models, 04.05.2023.
2. FTC, Generative AI Raises Competition Concerns, 29.06.2023.
3. European Parliament, Artificial Intelligence Act.
4. European Parliament, Artificial Intelligence Liability Directive.
5. Autorité de la concurrence, Generative artificial intelligence: the Autorité issues its opinion on the competitive functioning of the sector, 28.06.2024.
6. Audoridade da Concorrência, AdC warns of competition risks in the Generative Artificial Intelligence sector, 05.11.2023.
7. Hungarian Competition Authority, GVH launches market analysis on the impact of artificial intelligence, 04.01.2024.
8. CMA, Amazon / Anthropic partnership merger inquiry, 08.08.2024.
9. OECD, Artificial intelligence, data and competition - Background Note, 06.05.2024, para. 93, 94, 142.
10. Türkiye Büyük Millet Meclisi, Yapay Zeka Kanun Teklifi (Proposal for Artificial Intelligence Law), 24.06.2024.
The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances.