The Department of Justice and eight state attorneys general filed a civil antitrust lawsuit in August against RealPage Inc. – a Texas-based software company that provides property management software – and several landlords using its software. This case adds to the growing number of antitrust cases targeting algorithmic pricing tools, and is another example of federal regulators taking a tough stance on new AI technologies.

The DOJ’s Allegations

The complaint largely focuses on RealPage’s AI-trained pricing algorithm, which generates pricing recommendations for property owners and managers. According to the DOJ, RealPage’s algorithm leverages data from participating landlords and property management companies to generate pricing recommendations for units and properties. These pricing recommendations tend to be similar among comparable, participating properties. RealPage also contracts with landlords to receive competitively sensitive information about apartment rental rates and lease terms, which can further hone RealPage’s algorithmic recommendations regarding terms and conditions of different leases. The complaint asserts that this practice insulates prices from the competitive market, inflates rent, and shapes the terms of rental agreements in a way that harms hundreds of millions of tenants across the country. The DOJ alleges this violates Sections 1 and 2 of the Sherman Act. 

Separate from RealPage’s alleged impact on the multi-family rental building market, the complaint also alleges that RealPage’s actions violate the Sherman Act by monopolizing the commercial revenue management software market, where it holds an 80% share. Under the agreements between landlords and RealPage, landlords must use RealPage’s revenue management systems exclusively to give pricing quotes to potential renters. The complaint frames this as another grounds for violations under the Sherman Act Sections 1 and 2. 

The Evolving Enforcement Approach

The complaint sheds light on how the DOJ might construe certain AI-powered algorithms as facilitating collusive behavior among competitors. The old safe harbor guidelines allowed competitors to share information and benchmark in good faith. But those guidelines were withdrawn last year. The DOJ complaint seems to suggest that it might interpret any algorithm using nonpublic data to make recommendations as inherently anticompetitive.

However, in a notable shift, the DOJ is not pursuing a per se theory in the RealPage case, even though it has made this argument in statements of interest filed in prior private antitrust cases involving pricing algorithms. Instead, it is arguing that the rule of reason applies, which asks courts to weigh the anti-competitive effects of conduct against its pro-competitive justifications.

While the DOJ stops short of saying that the algorithm itself is indicative of a horizontal agreement among parties, it could argue that the algorithms should be construed as a “plus factor” weighting in favor of such a finding. Specifically, the DOJ might view the aggregation and subsequent sharing of data via RealPage’s AI model as “communications” among competitors, the equivalent the kinds of meetings or opportunity to agree on a conspiracy used in other case. Furthermore, the algorithm’s self-proclaimed resistance to market forces demonstrates provides a platform for the DOJ to argue how adherence to its recommendations contravenes a landlord’s economic interests to react to changes in the market. Finally, the infrastructure surrounding compliance with the algorithm – mainly RealPage’s own practice of ensuring properties adhere to its pricing recommendations – could be interpreted as providing motive to collude (another plus factor).

Parallel Private and Federal Proceedings

This complaint is not the only enforcement action taking aim at this algorithm. In the of spring 2023, a private action against RealPage and its clients was filed in the Middle District of Tennessee, containing similar allegations regarding the platform’s AI-trained algorithmic pricing model. That class action is currently in discovery, having survived motions to dismiss. There, the court found that RealPage’s statements that it allows the sharing of competitive price information among landlords to be strongly indicative, but not solely supportive, of the finding of a horizontal agreement between landlords.  Key instead for the court was the parallel conduct of defendants – that is, the raising of prices even if doing so resulted in higher vacancy rates (contravening market forces). An important difference between the private action against RealPage and the DOJ’s RealPage action is the focus on the role of the algorithm. The DOJ filed a statements of interest in the private case, urging a per se approach be used to scrutinize the role of the algorithm in coordinating pricing conduct, whereas in the DOJ’s recent action does not urge a per se approach. 

AI-Trained Pricing in and of Itself is Not the Issue

There are plenty of AI-powered pricing algorithms that would not necessarily run afoul of the DOJ’s theory here. Unlike some other AI pricing algorithms that businesses operate in-house and independently, RealPage’s software aggregates private data from multiple outside landlords and property management companies. By pooling this information, RealPage can generate pricing recommendations that reflect not just general, public market conditions but also the specific actions of other landlords using the platform. This is suggested in the DOJ complaint, which was filed in the U.S. District Court for the Middle District of North Carolina and focuses in part on rental pricing in the Chapel Hill region which has a robust rental population of students each semester.

Time will tell whether and how courts will distinguish RealPage’s price recommendations from other algorithmic AI price recommendation software, some of which have not survived motions to dismiss. For example, in Gibson v. Cendyn Group, a Nevada federal court found that plaintiffs failed to adequately allege that Las Vegas hotel operators violated Section 1 of the Sherman Act by agreeing to set hotel room prices using pricing algorithms from the same vendor, Cendyn. That ruling explicitly distinguished its facts from the RealPage case in the Middle District of Tennessee: “This case does not involve allegations of competitors pooling their confidential or proprietary information in the dataset that the pertinent algorithm runs on, while that case did” (emphasis added).

In Cornish-Adebiyi v. Caesars Entertainment, a New Jersey district court is considering defendants’ motion to dismiss a similar case against Atlantic City hotel operators and Cendyn, the same vendor as the Las Vegas-based Gibson case. But the Cornish complaint includes allegations that the algorithm at issue relies on nonpublic data provided by competitor subscribers – potentially aligning its outcome more closely with RealPage than with Gibson, assuming that the nonpublic nature of the pooled information is indeed the distinguishing factor. 

One additional potential distinguishing factor is the unique role that geography plays in the real estate housing market, given how renters are tied to a particular marker. This factor would be considerably less salient for hotel users than for long term renters.

Practical Impact

Separate from questions specific to the still-emerging legality of AI technologies, the DOJ’s RealPage complaint marks another interesting trend of DOJ enforcement actions following private actions. For example, the DOJ’s action against the NCAA followed a private action alleging similar claims. While it isn’t clear what to make of this trend just yet, it is a notable reversal from times when the DOJ or the FTC led the way for private actions to follow and obtain monetary relief. 

The DOJ’s two-track approach is also noteworthy. Reading its per se theory in the statement of interest filed in the private action and the rule of reason approach in its own action together, it seems like the agency is covering its bases to ensure some sort of Sherman Act violation is found.

While definitive legal theories are still emerging, any company reliant on AI-driven algorithmic pricing trained by a nonpublic, competitively sensitive data set should keep close watch on this case, which could have a massive impact on pricing models in a wide range of industries. In addition to antitrust concerns, there may be emerging legislative steps that might implicate their software.  San Francisco, for example, recently passed an ordinance that bans both the sale and use of software which uses non-public competitor data to set, recommend, or advise on rents and occupancy levels.

Companies using pricing algorithms should review how these tools work, document any procompetitive benefits, and check if they are insured or protected against investigations. There are many ways to minimize scrutiny in these situations – antitrust and otherwise – and the best solutions will be tailored to the specific tool and client.