After the chatbot model ChatGPT came into public view in 2022, the Internet was buzzing with talk about the book “The Singularity Is Near: When Humans Transcend Biology” that was popular worldwide nearly 20 years ago.  In that book, the author Ray Kurzweil, a renowned artificial intelligence (hereinafter referred to as “AI”) scientist and Google’s chief futurist, suggested that AI will completely surpass humans by 2045 and that humans and machines will begin to combine, with carbon-based life replaced by silicon-based life[1] .

The legal issues arising from AI in the antitrust field have been receiving increasing attention from regulators in various countries.  In particular, the huge amount of user data and the opaque and complex algorithms held by leading technology companies will increase the monopoly power of these technology companies and impede the effective competition companies in this sector.  This article briefly analyses the antitrust-related issues that may be caused by AI from three aspects,(i) monopoly agreements, (ii) abuse of market dominant position, and (iii) concentration of operators, thereby stimulating more exchanges and discussions.

1. Monopoly agreements

Articles 5 to 8 of the Anti-Monopoly Guidelines of the Anti-Monopoly Commission of the State Council for the Sector of Platform Economy (hereinafter referred to as the “Guidelines”), effective in February 2021, provide for horizontal algorithmic monopoly agreements, vertical algorithmic monopoly agreements, and hub-and-spoke algorithmic agreements.  Article 5 of the Guidelines stipulates that monopoly agreements in the platform economy refer to agreements, decisions or other concerted behaviours by operators that aim to exclude or restrict competition.  Agreements and decisions can be in written, oral or other forms.  Other concerted behaviours refer to behaviours that are essentially concerted through data, algorithms, platform rules or other means, although the operators have not explicitly entered into agreements or decisions, except for parallel behaviours such as price following based on independent expression of intent made by the operators concerned.  A similar position is held by foreign regulators that, with regard to pricing algorithms, if the algorithms between operators are independent of each other, no illegal liability will occur as long as there is no collusion to fix prices.

With the development of AI technology, more and more companies are using algorithms to help determine the prices of their products and services.  In the Meyer v. Kalanick case[2], Spencer Meyer, an Uber passenger, filed a class action against Travis Kalanick, CEO and co-founder of Uber, for planning and assisting Uber drivers with price fixing.

Spencer Meyer opined that there was a horizontal monopoly agreement among Uber drivers who fixed prices by using Uber’s pricing algorithm.  Although there was no direct price coordination among drivers, each driver using the algorithm knew that other drivers would use the same algorithm (Uber’s pricing algorithm produced prices above the competition level).  Moreover, in the absence of collusion among competitors, drivers would deviate from the established algorithm to charge lower prices, in order to compete for customers.  Travis Kalanick argued that each driver entering into a contract with Uber was an independent act, that a driver’s agreement to use Uber’s pricing algorithm has no impact on the independence of the driver’s decision to join Uber, and that drivers entered into agreements with Uber in order to take advantage of the payment and matchmaking services provided by Uber.  In the above case, the court did not accept Travis Kalanick’s arguments and held that a driver who agrees to use Uber’s pricing algorithm does so with the clear knowledge that all other drivers will also agree to use the same pricing algorithm.

Although the above is a case under U.S. law, the same case facts may also constitute a monopoly agreement under China’s current antitrust laws and regulations.

The current challenge facing antitrust enforcement agencies in various countries is the illegality of tacit algorithmic collusion (also known as self-learning algorithmic collusion).  Specifically, an algorithm is able to autonomously learn, in an unknown environment, how to achieve pre-specified goals, such as profit maximisation, without human intervention.  Such algorithm can learn how to implement collusion strategies based on reward and punishment schemes, ultimately leading to and maintaining supra-competitive prices. The algorithm itself is neither programmed for the purpose of collusion, nor is it biased to support the formation of monopoly agreements.  Instead, the algorithm adopts collusion strategies through autonomous decision-making.[3] Simply put, by virtue of the powerful predictive capability, an algorithm can autonomously arrive at a collusive outcome without human intervention by continuously learning and adapting to the behaviour of enterprises in the market.

2. Abuse of market dominant position

Article 22 of the Anti-Monopoly Law of the People’s Republic of China (hereinafter referred to as “AML”) prohibits an operator with a dominant market position from, without justifiable reasons, applying differential treatment in terms of transaction prices and other transaction conditions to counterparties with the same conditions.  In accordance with Article 21 of the Provisions on Administration of Algorithmic Recommendation in the Internet Information Services, differential treatment in algorithmic recommendation services is mostly about the use of algorithms by algorithmic recommendation service providers to implement unreasonable differential treatment in transaction prices and other transaction conditions based on consumer preferences, transaction habits and other characteristics.  The more common ones are algorithmic self-preferencing and algorithmic personalised pricing (or “big data discrimination”). 

Take algorithmic self-preferencing as an example.  In the Alibaba Group’s “either-or” monopoly case[4], the State Administration for Market Supervision (hereinafter referred to as “SAMR”) pointed out that the core of search algorithms is to improve the search conversion rate, so that goods get more attention from consumers, thereby increasing sales.  Alibaba lowered the search weight of some platform operators that did not implement the “either-or” requirements, which directly led to their goods being ranked lower on the platform, or even not being found at all.  Alibaba’s reason for manipulating the natural ranking results was to improve the search ranking of the goods of the platform operators who did implement the “either-or” requirements, which was suspected of constituting self-preferencing.  The purpose of self-preferencing is to increase Alibaba’s market share, or to impose algorithmic punishment on counterparties in order to prohibit or restrict counterparties from trading with its competitors.

3. Concentration of operators 

Concentration of operators under Article 25 of the AML includes (i) merger of operators, (ii) acquisition of control over other operators through equity/assets acquisition, and also (iii) acquisition of control or exertion of decisive influence over other operators through contracts or other means.

It has been observed that more and more enterprises in the AI sector are trying to avoid antitrust examination of mergers and acquisitions to some extent by means of establishing partnerships (such as data and computility partnerships, providing financing, and scouting of expert talents) or minority equity investments.  Therefore, antitrust regulators in various countries are increasingly paying attention to the antitrust examination of such partnerships, focusing on whether such partnerships will constitute a concentration of operators and monopoly behaviour.

In December 2023, the UK Competition and Markets Authority (an antitrust regulator) launched an investigation into Microsoft’s partnership with OpenAI, examining whether the partnership gave Microsoft de facto control over OpenAI or more than 50% of the voting rights, in order to determine whether their close and multifaceted relationship constituted a transaction that may have an impact on competition in the UK market and thus needs to be reported.[5] In September 2024, upon an antitrust investigation, the UK Competition and Markets Authority approved Microsoft’s hiring of Inflection AI’s key employees (including co-founders, AI engineers, researchers, etc.) and the acquisition of Inflection AI software licences for $650 million.  The UK Competition and Markets Authority opined that Microsoft’s hiring of Inflection’s core team plus the team’s related “know-how” was sufficient to constitute a takeover of the business[6].  This case reflects the expansion of the UK Competition and Markets Authority’s examination and its focus on partnerships in the AI sector.

In China, both the AML and the Regulations on the Examination of Concentration of Operators grant the SAMR the power to proactively intervene in the investigation of transactions that do not meet the reporting threshold, which will be the main legal basis in China for regulating concentrations of operators in the AI sector through mergers and acquisitions or partnerships. 

In addition, in accordance with the Guidelines, we are of the opinion that China’s antitrust regulator will mainly consider the following points when examining the concentration of operators[7]: 

(i) the operator’s capabilities to control and process data, as data collection and analysis based on algorithms and AI may result in enterprises with large amounts of data having the monopoly power that impedes competitors from accessing the market; 

(ii) the impact of further concentration of key resources (e.g. computing resources such as data, cloud computing services and dedicated chips, and top technical talents) on innovation and technological advancement; and 

(iii) whether monopoly or significant market power will be formed in the market for key elements such as data and computility.

As we know, Google has integrated its Gemini model into multiple Google services, Microsoft has deployed Copilot models in its Office 365 software, and Apple has established a technology partnership with OpenAI to deeply integrate ChatGPT into Apple’s product matrix.  With the AI models deployed in their products and services, large technology companies will further consolidate their dominant position in the market as companies obtaining more data can train more accurate models and provide better services, thereby attracting more users (who also tend to choose products and services with larger user bases), which results in the positive cycle of “data collection -→ algorithm optimisation -→ enhanced user stickiness -→ increased business value”.  In addition, the protection of patents and intellectual property rights and the self-reinforcing AI enable leading technology companies with complex AI technologies to build stronger technical barriers that may limit or hinder innovative impetuses and the development of competitors.

While enjoying the efficiency and innovation brought about by AI technology, we should also continue to keep an eye on the problems and challenges it may cause in the antitrust field, establish better legal and regulatory mechanisms to ensure fair competition in the market, encourage enterprises to strike a balance between innovation and compliance, and continuously stimulate the potential of the digital economy.  The sustainable development of AI can only be better promoted by taking into account technology and systems, efficiency and fairness.

For more information or further guidance in this area, please contact Grace Zhao or Dr. Sven-Michael Werner.

VISIT OUR COMPETITION LAW HOMEPAGE

[1] Guangzhou Daily - Towards sci-fi silicon-based life, can humans?

[2] See Meyer v. Kalanick, No. 1:2015cv09796 - Document 37 (S.D.N.Y. 2016) :: Justia

[3] Weimin Shen｜Antitrust Regulation of Tacit Algorithmic Collusion_Shanghai Observer

[4] See the Decision on Administrative Punishment (Guo Shi Jian Chu [2021] No. 28) by the State Administration of Market Supervision

[5] U.K. Regulators Review Microsoft's Partnership with OpenAI - WSJ

[6]  https://assets.publishing.service.gov.uk/media/6719ff5f549f63039436b3c8/__Full_text_decision__.pdf 

[7] See Article 20 of the Anti-Monopoly Commission of the State Council for the Sector of Platform Economy.