AI in Antitrust: Harnessing technology for cartel detection and enforcement

Competitors colluding to shield themselves from market competition is not a new phenomenon. The infamous Phoebus cartel (where lightbulb companies colluded to make lightbulb lives shorter) was born in a meeting room in Geneva more than a century ago.

Yet, cartels of today are not hatched in smoke-filled rooms, but bear the marks of the digital age. As cartels have gained in sophistication with time and technology, antitrust authorities have had to keep pace to try and catch them. The recent rise of AI technology marks another cycle of this never-ending cat-and-mouse game.

The recently released “Artificial Intelligence and Competition” market study by the Indian antitrust authority, the Competition Commission of India (CCI), dedicates substantial space to the risks of AI-assisted cartelization (“algorithmic collusion”). However, despite some general references to capacity building, the enforcement enhancing potential of AI, especially anti-cartel enforcement, remained unexplored.

Dawn of AI-assisted cartel-busting

While AI can be a cartel-booster (making real-time collusion possible without explicit coordination between human actors), it can also be a powerful cartel-buster tool. Today, such AI-assisted enforcement is not just a theory, but is already driving enforcement practice globally. AI tools enable antitrust authorities to monitor and analyse vast database of publicly available information (prices, investor reports, earning calls, etc.) to screen for potentially suspicious activity,

For example, last year, the European Commission launched “dawn raids" on the offices of several tyre manufacturers (Michelin, Pirelli, Goodyear, and Continental) on suspicion of price-fixing. This cartel inquiry wasn’t triggered by a whistleblower, leniency applicant, or a third-party complainant. The trigger, instead, was an AI-driven screening of the transcripts of these companies’ public statements and earnings calls. In July, the European Union General Court validated this use of AI-tools to help detect cartels.

Public communication, hidden risks

The European Commission’s AI tool was trained to spot patterns and flag sectors where competitors' public language showed an unusual level of alignment. The tyre sector tripped the AI tool’s internal alarms. Subsequent manual review of the underlying data seemed to back what AI had flagged. While what specific word patterns triggered the AI tool’s internal alarms is not publicly known, problematic statements included language alluding to a company’s desire to "pass on" rising raw material costs, stressing the need for "market discipline," etc.

In a nutshell, the AI-tools helped the European Commission detect such potentially anticompetitive ‘public signaling’, i.e. competitors aligning themselves via public statements without any explicit coordination.

AI and public disclosures

Such AI-driven cartel detection has strong implications for businesses, especially publicly-traded companies which have substantial public disclosure obligations. All public communications (not just earnings calls, but press releases, conferences, presentations, and even executive interviews) need to be made knowing the competition law risks they can carry.

Each such communication must be weighed for how it might be interpreted not just by investors, but by a regulator’s algorithm. While companies can better train their executives for the future, not much can be done for past public communications. Any sins of the past will lie dormant and be forever susceptible to discovery by a regulator’s cartel detecting AI-tool.

For antitrust regulators (like the CCI), AI can unlock a treasure-trove of enforcement intelligence. Powerful AI-tools trained and operating on companies’ public communications over the years can help detect high-risk sectors for cartelization.

Following such initial screening, the CCI can deploy investigative resources to manually analyse the underlying data to determine whether a suo motu investigation is warranted or not. Further, the very fact that the CCI has such AI-capabilities can boost the effectiveness of its leniency regime. Firms, which would have not taken the leniency route earlier, may change their mind given the raised likelihood of detection by the CCI.

AI and Anti-cartel enforcement

The market study recommends increasing CCI’s institutional capacity, and international cooperation, to deal with AI-driven competition risks. Such capacity building could entail developing such AI-driven cartel detection capabilities. This would also be a prime area for inter-jurisdiction cooperation.

The CCI could collaborate with other jurisdictional antitrust authorities (such as the European Commission, the UK Competition Markets Authority, etc,) to develop AI-tools for anti-cartel enforcement. Such AI-tools would, of course, need to be suitably Indian-ized to account for the Indian legal and market realities.

(Shashank Sharma is the Policy Lead at Axiom5 Law Chambers)