Authored By: MUKESH KUMAR Yadav
National Law College, Nepal
Introduction
Section 1 of the Sherman Act condemns every contract, combination or conspiracy in restraint of trade. It has never condemned a company for charging a high price on its own. That distinction, which has organised American antitrust law for more than a century, is now under pressure from software. On 29 July 2026 the Court of Appeals for the Third Circuit reinstated a price-fixing class action brought by guests of Atlantic City casino-hotels who alleged that their room rates had been set by a common pricing algorithm.” Eleven months earlier the Court of Appeals for the Ninth Circuit had affirmed the dismissal of a materially similar claim brought against Las Vegas hotels using software supplied by the same vendor. Two appellate courts, one product, opposite results.
The divergence is not accidental. It reflects a doctrinal framework built for a world in which competitors coordinate by talking to one another. Pricing algorithms permit competitors to align their conduct without ever doing so: each firm contracts separately with a vendor, each retains formal freedom to reject the output, and the coordination, if it exists at all, occurs inside a model that no participant fully controls.
This article argues that the courts have responded to that difficulty by converging on a single proxy for agreement, namely whether the algorithm is fed non-public competitor data, and that this proxy is an inadequate substitute for the statutory concept of concerted action. It is under-inclusive because self-leaming systems can produce supracompetitive alignment without any pooling of confidential information, and it misidentifies the true gravamen of the conduct, which is the delegation of independent pricing discretion to a common decision-maker. Part II sets out the governing framework. Part III examines the emerging case law. Part IV evaluates the data-provenance test critically. Part V proposes a delegation-based standard supported by mandatory auditability.
The Governing Legal Framework
The Agreement Requirement
Liability under Section 1 attaches only to concerted action. Unilateral conduct, however anticompetitive its effects, falls outside the provision, because Section 1 reaches only the joining of separate economic actors pursuing separate interests. A plaintiff must therefore offer evidence that tends to exclude the possibility of independent action. At the pleading stage, Twombly holds that allegations of parallel conduct alone will not survive dismissal; the complaint must contain enough factual matter to render an agreement plausible rather than merely conceivable.”
Agreement need not be express. Interstate Circuit established that a conspiracy may be inferred where a common actor extends the same invitation to competing firms, each firm knows that the others have received it, and each acts in a manner that would be against its interest absent the participation of the others. This hub-and-spoke reasoning is the doctrinal ancestor of every algorithmic pricing claim now before the federal courts.
The Applicable Standard of Review
Horizontal price fixing is unlawful per se, without inquiry into its reasonableness or effects. Whether algorithmic coordination attracts that treatment is unsettled. The Middle District of Tennessee, presiding over the consolidated private litigation against RealPage, dismissed the per se count and allowed the case to proceed only under the rule of reason, reasoning that courts lack sufficient experience with pricing algorithms to declare them categorically pernicious. The Western District of Washington expressly disagreed, holding that the use of novel technology to effectuate an otherwise classic horizontal conspiracy does not remove the conspiracy from per se condemnation.
Information Exchange
A parallel body of doctrine governs the exchange of competitively sensitive information. Container Corp held that the reciprocal exchange of current price information among competitors in a concentrated market may itself restrain trade,” and later authority has treated the currency, granularity and non-public character of the data, together with market concentration, as the decisive variables. It is this line of authority, rather than the law of agreement proper, that has come to dominate the algorithmic cases.
III. The Emerging Case Law
Gibson v Cendyn Group
In Gibson, guests of Las Vegas Strip hotels alleged that the hotels had licensed revenue-management software from Cendyn and that room rates rose in consequence. The Ninth Circuit, the first appellate court to address the question, affirmed dismissal. Independent decisions by competitors to purchase the same product do not establish an agreement among them, particularly where the software imposed no obligation to accept its recommendations and did not circulate confidential information between licensees. Two features of the opinion repay attention. The panel accepted that it would undoubtedly violate Section 1 for competing hotels to agree among themselves to abide by a third party’s pricing recommendations, and it added, in a footnote, that its analysis might change had the plaintiffs alleged that the vendor shared each hotel’s confidential information with the others.
Cornish-Adebiyi v Caesars Entertainment
The Third Circuit accepted precisely that invitation. In Cornish-Adebiyi, the plaintiffs alleged that Atlantic City casino-hotels continuously supplied current, non-public rate and occupancy data to the same vendor’s software, that the resulting recommendations were loaded directly into each hotel’s booking system, and that the hotels adopted those recommendations approximately ninety per cent of the time, departing from them only through a deliberate override process.” Taken together with parallel conduct, market structure and observed rate increases, the court held that these allegations plausibly supported an inference of a hub-and-spoke conspiracy and reversed the dismissal. The two decisions are therefore reconcilable on their pleadings, but only because the Third Circuit found the very allegation whose absence the Ninth Circuit had identified as dispositive.
The District Courts and the Enforcement Record
Below the appellate level the picture is more fragmented still. In Duffy v Yardi Systems, the court sustained a complaint on a per se theory, finding an agreement plausibly alleged in the landlords’ acceptance of the vendor’s invitation to exchange sensitive information in return for higher rents.” A California state court, by contrast, granted summary judgment to the same vendor after source-code evidence showed that customers used the software independently and without contributing sensitive data. On the public enforcement side, the Department of Justice sued RealPage in 2024 and settled in November 2025 without any adjudication of illegality, accepting instead a set of design constraints: a twelve-month minimum age for training data, a prohibition on real-time lease inputs, a bar on geographic modelling below state level, and a court-appointed monitor with access to the company’s code and model documentation. A consent decree with a property manager followed in July 2026. The pattern that emerges is one of intense scrutiny producing no binding statement of the substantive rule.
Critical Evaluation: The Limits of the Data-Provenance Test
Read together, these authorities establish a working rule: algorithmic pricing is lawful unless the algorithm ingests non-public competitor data. The attraction of that rule is obvious. It is administrable, it maps onto the settled law of information exchange, and it avoids condemning dynamic pricing, which is ubiquitous and frequently procompetitive. It is nonetheless the wrong test, for three reasons.
First, it is under-inclusive. Economic modelling of reinforcement-learning agents indicates that independently operated algorithms, trained only on publicly observable posted prices, can learn to sustain supracompetitive equilibria through reciprocal punishment strategies, without communication and without pooled data. Under Gibson, such systems are lawful: there is no exchange of confidential information, no agreement to adhere, and nothing but parallel conduct to plead.” Yet the consumer harm is identical to that in Cornish-Adebiyi. A doctrine whose coverage turns on data provenance will lose purchase precisely as the technology it governs becomes more capable, because the more sophisticated the model, the less it needs its rivals’ secrets.
Second, the test confuses evidence with the element it is meant to prove. Data pooling is probative of agreement because firms do not ordinarily surrender confidential information to competitors without expecting something in return. But the substance of the wrong is the surrender of independent pricing discretion to a common decision-maker. The Ninth Circuit itself conceded as much when it accepted that an agreement to abide by a third party’s recommendations would undoubtedly offend Section 1.The facts that carried the day in the Third Circuit, namely a ninety per cent adherence rate and an override process that made deviation costly, are facts about delegation, not about data. The courts have identified the correct indicium and then filed it under the wrong heading.
Third, the resulting law is unstable across fora. Identical conduct attracts per se treatment in Washington and rule-of-reason treatment in Tennessee, with dramatic consequences for the plaintiff’s burden. Legislatures have begun to fill the vacuum unevenly: New York has amended the Donnelly Act to reach the use of pricing algorithms by residential landlords,” while California has amended the Cartwright Act to address common pricing algorithms and to relax the pleading standard for collusion claims generally.” Blanket prohibitions of this kind invite constitutional challenge, as RealPage’s First Amendment action against the New York statute demonstrates.
It is fair to object that this analysis overstates the danger. The Department of Justice, having pleaded an aggressive theory, chose to settle rather than litigate it, and the settlement leaves ordinary algorithmic pricing untouched. On that reading, enforcement has been calibrated rather than hostile, and further intervention risks deterring efficient price adjustment that benefits consumers in periods of low demand. The objection has force, but it is an argument for precision, not for inaction. The proposals below do not treat algorithmic pricing as inherently suspect; they identify the narrow class of arrangements in which pricing discretion has in substance been pooled.
Proposals for Reform
First, the concept of agreement should be operationalised through delegation rather than data. Congress should create a rebuttable presumption of concerted action where competitors accounting for a substantial share of a defined market adopt a common pricing engine and either contribute non-public inputs to it or accept its recommendations at a rate, and under override conditions, indicating that independent judgment has been abandoned. The Preventing Algorithmic Collusion Act, which triggers on non-public competitor data alone,” reproduces the defect identified above and should be amended to include an adherence-based trigger.
Second, the opacity problem should be addressed directly. Vendors of shared pricing tools should be required to retain model training documentation and runtime decision logs for a defined period and to produce them to enforcers on request. The monitoring provisions of the RealPage settlement supply a workable template, and a general obligation would render tractable a question that presently defeats most plaintiffs at the pleading stage: what the model actually did, and on what inputs.
Third, the standard of review should follow the substance. Where delegation is established, the conduct is horizontal price fixing effectuated through an intermediary and per se treatment is appropriate. Where firms use algorithms independently, the rule of reason should continue to govern. A federal standard on these lines would also displace the emerging state patchwork and confine regulation to conduct rather than to the vendor’s recommendations themselves, thereby avoiding the First Amendment difficulties that categorical prohibitions have already generated.
Conclusion
The conflict between the Ninth and Third Circuits is not a disagreement about the meaning of Section 1. It is a symptom of a doctrine that must locate an agreement in conduct that no longer requires one. The courts have responded pragmatically, treating the presence of pooled non-public data as the marker of concerted action, and that response has produced defensible outcomes on the facts before them.
This article has argued that the marker cannot bear the weight now placed upon it. It will fail against algorithms that require no confidential inputs to converge on supracompetitive prices, and it obscures the feature that actually distinguishes lawful dynamic pricing from unlawful coordination, namely the delegation of pricing discretion to a decision-maker shared with one’s competitors. A delegation-based presumption, coupled with mandatory auditability and a standard of review that follows the substance of the arrangement, would restore the fit between the statutory prohibition and the conduct it exists to reach.
The alternative is a body of law that grows more confident in its formulations as it becomes less capable of applying them. Competitors who once needed a room and a handshake now need only a shared subscription. The law should be indifferent to which of the two they choose.
Bibliography
Table of Cases
Bell Atlantic Corp v Twombly 550 US 544 (2007)
Copperweld Corp v Independence Tube Corp 467 US 752 (1984)
Cornish-Adebiyi v Caesars Entertainment Inc, No 24-3006 (3d Cir, 29 July 2026)
Duffy v Yardi Systems Inc, No 2:23-cv-01391-RSL, 2024 WL 4980771 (WD Wash, 4 December 2024)
Gibson v Cendyn Group LLC 148 F 4th 1069 (9th Cir 2025)
In re RealPage Inc, Rental Software Antitrust Litigation (No II), No 3:23-md-3071 (MD Tenn)
Interstate Circuit Inc v United States 306 US 208 (1939)
Mach v Yardi Systems Inc, No 24-063117 (Cal Super Ct, 6 October 2025)
Monsanto Co v Spray-Rite Service Corp 465 US 752 (1984)
RealPage Inc v James, complaint filed in the Southern District of New York, 25 November 2025
Todd v Exxon Corp 275 F 3d 191 (2d Cir 2001)
United States v Container Corp of America 393 US 333 (1969)
United States v RealPage Inc, No 1:24-cv-00710 (MDNC)
United States v Socony-Vacuum Oil Co 310 US 150 (1940)
Table of Legislation
Assembly Bill 325 (Cal 2025), amending the Cartwright Act, Cal Bus and Prof Code s 16720 et seq
Assembly Bill 1417-B / S 7882 (NY 2025), amending the Donnelly Act, NY Gen Bus Law art 22
Preventing Algorithmic Collusion Act of 2025, S 232, 119th Cong (2025)
Sherman Antitrust Act 1890, 15 USC ss 1-2
Secondary Sources
Baker McKenzie, ‘United States: Ninth Circuit on Algorithmic Pricing’ (Baker McKenzie InsightPlus, 26 August 2025) https://www.bakermckenzie.com accessed 5 September 2026
Covington and Burling LLP, ‘Third Circuit Reinstates Algorithmic Pricing Antitrust Lawsuit’
(Covington, August 2026) https://www.cov.com accessed 5 September 2026
Davis Polk, ‘New Laws Regulating Algorithmic Pricing Enacted in New York and California’
(Davis Polk Client Update, 27 October 2025)
https://www.davispolk.com accessed 5 September 2026
Mayer Brown, ‘Algorithmic Price Fixing After Cornish-Adebiyi’ (Mayer Brown Insights, July 2026) https://www.mayerbrown.com accessed 5 September 2026
Reed Smith, ‘Algorithmic Pricing under Pressure: DOJ’s RealPage Settlement Changes the Rules for Rental Markets’ (Reed Smith Viewpoints, 3 December 2025) https://www.reedsmith.com accessed 5 September 2026
Wilson Sonsini, ‘2026 Antitrust Year in Preview: Algorithmic Pricing’ (WSGR, 22 January 2026) https://www.wsgr.com accessed 5 September 2026
