Amishreya Gupta and Srishti Gupta
The ubiquitous influence of Artificial Intelligence in our mundane lives has inevitably triggered a long trail of incessant questions; especially regarding the potential impact of AI on the market competition. Collusion refers to combinations, conspiracies or agreements among sellers to raise or fix prices and to reduce output to increase profits.[i] In the current market structure more often than not prices are set by machines since these machines possess certain features that allow them to achieve supra-competitive pricing without any contact; without the presence of any agreement between parties. In the archaic scenario, the whole idea of collusion relied on the agreement between corporations run by humans.
Whether AI and market competition share a symbiotic relationship with each other and whether the markets with the help of AI are potent enough to distort the market competition (via collusion) has raised many eyebrows in recent years. The author endeavours to examine the effect of AI and if the current legal instruments suffice to deal with emerging opportunities and challenges.
Understanding AI and Algorithms:
In terms of computation methods, an algorithm is a tool that can be used to solve a variety of complex problems involving data and calculations. One of the best illustrations to depict the role of algorithms in our daily lives is the use of Google as a search engine. To provide us with accurate information, Google uses algorithms to extract relevant information from the vast amount of data available in its domain. It takes one click and a fraction of seconds to extrapolate all the relevant information that we can access from our phones, laptops etc.
AI-enabled computers are capable of performing tasks that a human mind would not be able to. It is capable of processing large amounts of data in a matter of seconds, and with the added capability of self-learning and responding appropriately, it has become a must-have.
AI and Collusion: a new challenge for competition regulation:
The incredible augmentation in the automation of computerized procedure and the sharp advancement in the technology sector have changed the way we intermingle, talk and trade.[ii] With the world becoming more digitalized algorithms and AI have emerged as the big players of this game; players in the modern business landscape can monitor customer behaviour, collect data, and respond in real-time. Sophisticated data and algorithms can be used to target the consumer via personalized and behavioural advertisement thereby resulting in discriminatory pricing.[iii] The proliferation in the number of e-users has also resulted in a boost in the data-driven economy. Players in the digital space can deliver a wide variety of creative and customised services thanks to business models that rely on massive data collection and processing in near real-time.[iv]
Companies these days are greatly relying on algorithm-based technologies to keep up with price pattern against their rivals. The firms’ pricing algorithms may result in collusive outcomes.[v] Algorithmic collusion, in essence, refers to a case in which algorithms are used in some way to help businesses collude with one another. In simpler terms, it involves using pricing algorithms to collude with rivals without the need for human interference. These pricing algorithms are a way of manipulating data and can aid in the facilitation of explicit firm cooperation agreements, and under some circumstances, they can even contribute to Tacit Collusion.[vi]
Tacit Collusion occurs when there is no explicit cooperation or agreement between firms, but they recognise competitive behaviour in the market and adjust their strategies accordingly; this type of collusion does not require the companies to declare their willingness to participate in the price-fixing agreement. It is a self-contained method that allows businesses to determine the most advantageous line of attack. In crux, businesses depend on self-learning machines to run their operations, and they have no idea about price-fixing or cartel persistence.[vii] Further, this category of collusion offers no room for the competition authoritiesdue to the lack of sufficient tools at their disposal. The issues here may go beyond the ambit and scope of antitrust laws because of the complexity of deciding the accountability of humans for the behaviour of a machine.[viii] Artificial intelligence-enabled collusion seems impeccable; it involves no agreement, intent, concerted practice or any conscious parallelism and because of the presence of nothing; it attracts nothing from the antitrust act.
AI and Collusion: Regulations around the World:
Article 81 of the European Commission Treaty prohibits agreements between undertakings, decisions by associations of undertakings and concerted practices which may affect trade between member states and which have as their object or effect the prevention, restriction or distortion of competition within the common market.[ix] To attract the provisions of Article 81, it is sufficient that the parties have expressed their intention to adopt specific conduct on the market.[x] Here the concept of ‘agreement’ revolves around the concurrence of wills between at least two undertakings.[xi] This means that the unilateral conduct does not fall within the ambit of Article 81.
In Polypropylene[xii], the EC held that an agreement exists merely “if the parties reach a consensus on a plan that limits or is likely to limit their commercial freedom by determining the lines of their mutual action or abstention from action in the market”.[xiii] Thus, even ‘inchoate understanding’, ‘conditional agreements’ and ‘vague and loose arrangements’ also falls within the scope of Article 81.[xiv] Basically, under the EC law, the existence of ‘expression of joint intention’ or the ‘concurrence of will’ between at least two undertakings is required.
Section 1 of the Sherman Act, 1890 confines itself to the joint actions of the parties involving “contract, combination or conspiracy” in restraint of trade.[xv] For most kinds of anticompetitive business conduct, condemnation has depended and continues to depend on finding two or more parties who may be said to have ‘agreed’ to do what was done, since ‘agreement’ is an essential ingredient of contract, combination or conspiracy.”[xvi] Under Sherman law, even after having wide scope in regard to the formation of the agreement it still requires an element of intent between at least two parties.
On the other hand, Section 5 of the FTC Actis more of a consumer protection Act and empowers the Commission to prevent unfair methods dampening competition within the state of America irrespective of the fact that whether it falls within the definition of the ‘agreement. If the action of any person/corporation is hampering the competition in the market, it is sufficient to attract the provisions of the Act.
Section 3 of the Competition Act of 2002 prohibits anti-competitive agreements between enterprises. The enterprises are prohibited from entering into an anti-competitive arrangement that has an appreciable adverse effect on competition within India under Section 3(1).
Section 3(3) expressly forbids collusion between enterprises and states:
“Any agreement entered into between enterprises or associations of enterprises or persons or associations of persons or between any person and enterprise or practice carried on, or decision taken by, any association of enterprises or association of persons, including cartels, engaged in identical or similar trade of goods or provision of services, which –
a. directly or indirectly determines purchase or sale prices;
b. limits or controls production, supply, markets, technical development, investment or provision of services;
c. shares the market or source of production or provision of services by way of allocation of the geographical area of the market, or type of goods or services, or number of customers in the market or any other similar way;
d. directly or indirectly results in bid-rigging or collusive bidding, shall be presumed to have an appreciable adverse effect on competition:”[xvii]
Our current collusion developments catechize questions on whether our current Act is capable enough of targeting and punishing collusion between machines or if it is inclined towards a narrow approach limited to humans. Without a shadow of a doubt, the problems of authorities have certainly exacerbated due to the current scenario; not only are the authorities finding it more difficult to uncover anticompetitive activity but in many cases, despite recognising anticompetitive behaviour they lack the necessary resources and provisions to address the problem.
Competition regulation usually focuses on possible illegal agreements between rivals, anticompetitive vertical constraints (such as resale price maintenance), abuse of dominant market power, and mergers with the potential to significantly stifle competition where human intervention and involvement are more like a prerequisite.
One key example of this would be the case of Samir Agarwal V/s Competition Commission of India & Ors[xviii] in which allegations of collusion inter se the drivers through the platform of Ola and Uber was held meritless. The existence of an agreement, understanding or arrangement, demonstrating or indicating the meeting of minds, is a sine qua non for establishing a contravention under Section 3 of the Competition Act, 2002; the use of algorithmically determined prices by the platform (Ola/Uber) could not be said to be amounting to collusion between the drivers. To fall under the ambit of a hub and spoke cartel an agreement between all drivers to set prices through the platform, or an agreement for the platform to coordinate prices between them is essential and since no such agreement between the drivers existed, the Commission did not find any violation of section 3(3) of the Act.
Under the Indian competition law, the term ‘agreement’ has an altogether different meaning and scope in comparison to the term ‘agreement’ under the Indian Contract Act, 1872.
The term ‘agreement’ as under Section 2(b) of The Competition Act states: “agreement” includes any arrangement or understanding or action in concert,—
(i) whether or not, such arrangement, understanding or action is formal or in writing; or
(ii) whether or not such arrangement, understanding or action is intended to be enforceable by legal proceedings;
The word “agreement” has a very broad meaning in this context; it encompasses any kind of arrangement, understanding or action, whether written or not, and the nature of the action is irrelevant for this Act. The Indian competition law, unfortunately, lacks relevant provisions to deal with collusion that involve self-learning machines.
Section 3, 2(b) of the Indian Competition Act 2002 needs an element of intent or some kind of agreement to attract these provisions without which this situation is less likely to improve. With no liability regime in place to ensure transparency for the phase of the complex algorithm, these challenges have the potential to threaten India’s competitive regime.
Both the EC and Indian competition law lack provisions to deal with collusion involving self-learning machines Article 81 of EC and Section 3, 2(b) of the Indian Competition Act requires an element of intent to attract the said provisions. Whereas, the only prerequisite under Section 5 of the FTC Act is the existence of anticompetitive activity; this signifies that firms using algorithms and AI-enabled machines (though unaware of the effects) will also be held liable under US law.
DRAFT COMPETITION (AMENDMENT) BILL, 2020:
The Indian Ministry of Corporate Affairs had set up an independent committee called, the Competition Law Review Committee (CLRC) to examine the Act along with accompanying rules and regulations. After consideration of the CLRC Report, the MCA has now prepared the draft Competition (Amendment) Bill, 2020 (Bill).
The Draft Competition (Amendment) Bill 2020 has suggested some fundamental changes which include but is not limited to the inclusion of ‘hub and spoke cartels’ within the ambit of ‘cartels’. This means an enterprise will now be presumed to be part of a horizontal agreement if it actively participates in the furtherance of the agreement, irrespective of whether it competes with other parties or not. Such inclusion will penalize hub and spoke cartels and shall expectantly help address tacit price collusion.
European Union and AI Regulations:
The European Commission on the recommendations of the European parliament, on April 21, 2021developed its proposal for an AI legal framework. This framework follows a risk-based approach and differentiates the uses of AI according to whether they create an unacceptable risk, a high risk, or a low risk. Companies developing and using high-risk A.I. applications such as self-driving software will be required to provide proof of safety and documentation explaining how the technology makes decisions. The companies must also guarantee human oversight in how the applications are created and used. Software-generated media content, including “deep fake” videos, will be subject to strict transparency disclosure. The creators must notify their users when the content is generated through automated means.
The proposed legal framework determines an A.I. application’s level of risk based on criteria including intended purpose, the number of potentially affected people, and the irreversibility of harm. Under the proposal, companies violating the rules could face fines of up to 6 percent of their annual global revenue.
Artificial intelligence has seen a momentous growth; the AI sector particularly has seen a significant upsurge in its value due to new technological advancements. This sector has managed to seep into all sectors including the competition regime. When it comes to cartels everything happens in the dark which makes the enforcement of competition law in such situations a mammoth task. The competition authorities must impart the knowledge of programming and algorithms among their people so to find out the real purpose of using the algorithm. Further, the authorities must realize that the present definition of ‘agreement’ has become obsolete and in the light of new technological advancement, it needs to widen its horizon. Moreover, the authorities should step in to regulate the type of information being exchanged among competitors and should limit it to information that is beneficial to the interest of consumers.
[i] OECD Glossary https://stats.oecd.org/glossary/detail.asp?ID=3159
[ii] Ezrachi&sStucke, Artificial Intelligence & Collusion: When Computers Inhibit Competition, 18 Oxf J Leg Stud 267, (2015).
[iii] Algorithm, Oxford dictionary,(Indian edition, 2007).
[iv]Ariel Ezrachi & Maurice E. Stucke, Virtual Competition(Harvard University Press, 2016).
[vii]Abhijeet Vikram Singh, Introduction of Algorithm and Artificial Intelligence into the Competition Law Regime: A New Market Dynamics, (2016). (LLM Thesis, NLU Delhi), http://hdl.handle.net/123456789/258.
[x] SA Hercules chemicals NV v Commission of the European Communities, (T-7/89) EU: T:1991:75.
[xi] Bayer AG v Commission, 2000 ECR.
[xii]DSM NV v the Commission (Polypropylene)  ECR II-1833.
[xv] Donald F. Turner, The Definition of Agreement under the Sherman Act: Conscious Parallelism and Refusals to Deal, 75(4), Harv. L. Rev. 655, (1962).
[xvii] Competition Act, 2002, No. 12, Acts of Parliament, 2003(India).
[xviii] Samir Agarwal v. Competition Commission of India & Ors. Competition Appeal (AT) No. 11/ 2019.
Any queries can be addressed via mail at firstname.lastname@example.org (Kindly mention “Query – Blog) at the mail.