The legal challenges raised by AI-powered algorithmic pricing

Artificial intelligence is reshaping the marketplace in ways most consumers are oblivious to. 

Through AI and algorithms, businesses can now adjust prices in real time based on market conditions and user data, including who you are, where you are, what you browse, and what companies think you are willing to pay. This often happens without the consumers’ knowledge or consent. Some retail apps even change their prices depending on whether you are inside or outside the store.

Algorithmic pricing has existed since the 1980s, when the airline industry set seat pricing based on supply and demand. However, its usage is rapidly becoming increasingly prevalent with advances in big data, AI, machine learning, real-time analytics, and cloud computing. 

Not surprisingly, businesses are embracing AI-powered algorithmic pricing in a range of sectors, including e-commerce, housing, ride-sharing, travel, and hospitality. 

Proponents maintain that it allows them to react swiftly to market changes, broaden market coverage, improve efficiency, provide a competitive advantage in fast-paced markets through personalized offers, and boost revenue and profits. 

“Generally, pricing algorithms are economically beneficial as they intensify competition, incentivize innovation and increase availability of goods and services,” Liam MacDonald, the director of policy and government relations at the Canadian Chamber of Commerce, said in a recent submission to the Competition Bureau after the federal agency released a discussion paper on algorithmic pricing.

Regulators keeping a watchful eye

However, as algorithmic pricing gains traction, it is under mounting scrutiny from government regulators and private litigators. They’re keeping a watchful eye on potential opportunities and risks for new and subtle forms of anti-competitive behaviour, consumer harm and privacy risks. 

“It definitely should be a top priority for regulators as well as the Competition Bureau, given the sophistication level that’s enabled by AI and its ability to influence markets and consumer behaviour,” says Irma Shaboian, an associate in the competition and foreign investment group at Stikeman Elliott LLP in Toronto. 

“It raises new challenges for competition enforcement, consumer protection, and privacy.”

For one, using AI in pricing systems could facilitate explicit price-fixing arrangements among competitors without leaving a paper trail. Just as disconcertingly, algorithms can be a catalyst in hub-and-spoke agreements, where several companies use the same third-party pricing algorithm software for pricing recommendations without explicit communication between them. According to the Competition Bureau, at least 60 companies in Canada offer price optimization software services. 

“It seems like companies are moving from the smoky room discussions to discussions using AI and using third parties to do the agreements for them,” says Linda Visser, a class action lawyer with Siskinds specializing in price-fixing conspiracies.

The real estate sector has become a prime target. The federal Competition Bureau is investigating whether corporate landlords are coordinating illegal rent hikes using AI algorithmic pricing software by American-based RealPage Inc. 

The move comes after the U.S. Department of Justice (DOJ) filed a civil antitrust suit last year against the company for its “unlawful scheme” to decrease competition among landlords in apartment pricing.

This summer, the DOJ reached a tentative settlement with Greystar Management Services LLC, the largest American landlord, that would bar it from using algorithmic rent-setting software that generates pricing recommendations using competitors’ data. Estimates are that the “anticompetitive” algorithmic rent pricing cost tenants US$3.8 billion ($70 per tenant each month) in 2023.

Concerns about algorithmic collusion

Perhaps even more unnerving is the eerie notion of algorithmic autonomous tacit collusion. A growing body of research suggests that self-learning autonomous algorithms can learn to collude without explicit programming or human intervention. According to a recent paper published by the Montreal-based Center for Interuniversity Research and Analysis on Organizations, the algorithms learn that keeping prices high generates better profits for all, and competing on price only triggers a price war “with no lasting gain.”

A 2024 study of gas stations in Germany found that the profit margins of those using an AI pricing algorithm rose by approximately five per cent. As more stations adopted the same pricing tool, prices “stabilized” at higher levels. 

“Where I think it really gets tricky, to be completely blunt and honest, is when you get to the place where artificial intelligence allows for kind of autonomous learning by the tools and adjustment of pricing strategies, adjustment of how firms compete, and potentially adjustment in terms of if or how they communicate with other tools,” says Trevor MacKay, deputy commissioner of the digital enforcement and intelligence branch at the Competition Bureau.

“That’s where it becomes an even bigger challenge for us to dig in and understand what's going on.”

The reality is, regulators are caught in a bind. The determination of the competitive or anti-competitive effects of algorithmic pricing, particularly personalized ones, hinges in large part on the goals that drive competition law and policy, says Pascale Chapdelaine, a law professor at the University of Windsor and co-founder of the Law and Technology Lab.

If the goal is to increase “consumer welfare,” she says these pricing strategies deserve closer scrutiny as potentially anti-competitive. But if the objective of competition law and policy is to boost overall economic or social welfare, as is arguably the case in Canada, then consumer harm stemming from personalized algorithmic pricing could be offset by the increased revenues of companies, with more or less “neutral effects” on the overall social or economic welfare. 

“The problem with this approach is that we're talking about prices that are set by gathering personal information about us without our knowledge, and that is something I don't think is fair,” Chapdelaine says. 

“People don't know the extent to which this is happening. That is a major problem.”

American authorities have taken a hard line against algorithmic collusion. 

“If we do not take a strong stand now … we will see this new form of price fixing destroying effective competition across a whole range of digital markets,” Roger Alford, the former principal deputy assistant attorney general in the DOJ’s antitrust division, said recently. 

On this side of the border, the Competition Bureau favours a balanced approach between fostering innovation and protecting competition, signalling that algorithmic pricing violations will be pursued as vigorously as traditional cartels. 

“We're in a situation where we've got an economy that we're looking to jumpstart,” says MacKay. 

“Productivity has not been where we want it to be over the last many years, and we see AI as an emerging technology that has great potential to be a productivity-enhancing tool that can be leveraged by firms, including small and medium enterprises, to gain access and compete … in a wide range of markets across the economy. So it's an important tool.”  

At the same time, he recognizes that algorithmic pricing can “enhance” anti-competitive conduct and practices that will get the Bureau’s attention, which they’ll have to be on top of. At the best of times, these practices “are going to test the bounds of our current laws.”

Current regulatory framework priced out?

Besides wondering whether regulators have the necessary technical expertise to effectively probe these pricing software tools, given the ‘black box' nature and opaqueness of AI-driven algorithms, coupled with companies' reluctance to share their intellectual property, there is a healthy debate within legal circles as to whether the current regulatory framework can address the novel issues raised by algorithmic pricing.

The Organisation for Economic Co-operation and Development says that existing competition law “sufficiently captures” cases where an algorithm facilitates explicit collusive agreements between humans, but not algorithmic autonomous tacit collusion. The international organization suggests reviewing the legal standards for “agreement” and “concerted practice” and moving away from relying on the notions of “act of reciprocal communication between firms” or “meeting the minds” as perquisites for antitrust intervention. 

Closer to home, in a joint submission to the Competition Bureau last year, seven Canadian class action law firms recommended that Canada emulate proposed American legislation currently in the Senate. Introduced last year, the Preventing Algorithmic Collusion Act creates a presumption of illegal agreement when direct competitors share competitively sensitive information through a pricing algorithm. It also compels companies to disclose the use of algorithms to set or recommend prices. 

The law firms also suggested Canadian regulators consider the European Union’s Artificial Intelligence Act as a “helpful” model. Under the Act, which is now in its implementation phase, national competition authorities and other designated regulators have new powers to access company information relating to AI systems. 

“Firms using AI-driven pricing algorithms should be required to disclose the functioning of the systems to regulators,” the law firms said. 

“Further, because many AI systems are ‘black boxes’ that may not have predictable or explainable responses, regulators should have powers to test AI pricing algorithms under controlled conditions in ‘regulatory sandboxes’ to observe how AI systems function in competitive environments.”

A dystopian coupling

In spite of calls for legal reforms, pundits maintain that traditional competition laws and enforcement mechanisms can effectively address the majority of the legal issues raised by algorithmic pricing. 

“I would love it if the law were clearer, but I think the laws still capture misconduct,” says Visser. 

“Companies are exchanging confidential information through the algorithm, knowing that their competitors are also doing so, and knowing that the result is a recommended price that gets spit out to them. So I would say that is a conspiracy that falls within the existing law.” 

Establishing proof of collusion in algorithmic pricing cases may be more “straightforward” than traditional conspiracies where you have to prove what the communications were, what exactly was said, and when the parties agreed to participate, says Visser, who believes AI algorithmic pricing will prove to be fertile ground for class actions.

The big question is how the courts will interpret the novel questions that will inevitably surface from algorithmic pricing, particularly those employing AI. 

“Where we'll end up in terms of how the courts interpret the ones that really stretch the bounds of our laws is an open question, but we'll only get those answers with time,” says MacKay. 

Shaboian says current legal tests weren’t designed with AI in mind.

For instance, what’s an agreement, she asks rhetorically. Can there be an agreement on the use of software? When did there become an agreement? When did the parties agree? 

“With traditional cases, it's still quite tough to pinpoint,” she says. 

“And with algorithmic AI, that would be perhaps even more difficult.” 

Chapdelaine says an auspicious avenue to tackle algorithmic pricing would be to use privacy laws. It’s a tactic legal experts believe holds promise. The use of a potential customer's personal information to assess their maximum willingness to pay goes against the core principles of valid consent and reasonable purpose under personal data protection law, she says. That may give rise to supplier breach of consumer contracts, such as a breach of general express warranties where a supplier is supposed to abide by privacy laws. 

“Once it’s determined that there's a violation of personal data protection law, then there’s a stronger argument to be made that there’s a violation of consumer protection law,” she says.

These are indeed heady times for regulators and the legal profession. The dystopian coupling of AI and algorithmic pricing raises legal issues that are only beginning to be dealt with. Lawyers counting on the Competition Bureau to provide guidance on how it will enforce or interpret the law will be disappointed, as none is forthcoming. 

In the meantime, lawyers must at least understand how AI works and be patient until the courts shed some light. 

“This brings a new challenge,” Visser says.