Shared Pricing Software Just Became an Antitrust Risk

Photo by Gorilla ROI Data Connector on Unsplash

The Vendor Contract Nobody Reads Until Discovery

A revenue manager at a 400-unit apartment operator logs into a pricing dashboard on a Tuesday morning, sees a recommended rent of $1,845 for a two-bedroom, and clicks accept. Nothing about that click feels like a conspiracy. It feels like using productivity software. That gap — between how a tool feels to the operator and how it looks to a plaintiff's antitrust lawyer — is the whole story of Cornish-Adebiyi.

According to Google News coverage of a Mayer Brown insights publication, the Third Circuit Court of Appeals revived hub-and-spoke conspiracy claims against shared pricing software in the Cornish-Adebiyi litigation, reversing a lower court's dismissal and allowing antitrust claims to proceed. The underlying allegation: landlords used shared pricing algorithms to coordinate rent increases. The legal architecture: the software vendor is the "hub," the participating landlords are the "spokes," and the shared algorithm is the rim that ties them into a single conspiracy.

The bolded thesis for anyone who buys software: the antitrust exposure in a shared-data pricing tool does not come from what the tool does for you — it comes from what the tool does with your competitors' data, and that is a procurement question, not a legal one.

The Job You Are Actually Hiring Pricing Software To Do

Strip the category down to the job-to-be-done and it splits cleanly in two, and almost every buyer conflates them.

Job A: read my own demand signal faster than a spreadsheet can. Occupancy, lease expirations, seasonality, historical conversion rates, days-on-market for your own units. This is internal-data optimization. It is genuinely hard, genuinely valuable, and — critically — it involves no competitor information at all.

Job B: tell me what everyone else is charging so I do not leave money on the table. This is the job that sells. It is also the job that puts a vendor in the middle of a hub-and-spoke diagram. As of August 4, 2026, based on the reporting available on Cornish-Adebiyi, it is Job B that the Third Circuit found plausibly supports a conspiracy claim — not the software's existence, but the pooling of non-public competitive inputs and the return of a coordinated output.

The non-obvious point the surface coverage tends to skip: most buyers never asked which job they were buying. The demo shows both fused into one screen. There is no toggle in the sales deck labeled "use competitor data: on/off," and in many deployments there is no such toggle in the product either. The moment you outgrow a spreadsheet, the market pushes you toward tools that solved Job B first and bolted Job A on top, because competitive data is the harder moat to build.

Where the "It's Just Software" Defense Breaks Down

The instinctive defense is intuitive and, on its own terms, reasonable: no two landlords ever spoke. There was no smoke-filled room, no email, no handshake. Traditional antitrust doctrine was built for human conspirators who agreed to something. A recommendation engine is not a person.

Here is where a careful skeptic should push back — and where the pushback fails. The hub-and-spoke theory was never about direct spoke-to-spoke contact. It has always been about whether the spokes knowingly participated in a scheme they understood their rivals were also joining. Substitute an API (a way for two apps to exchange data automatically) for a phone call and the doctrinal question barely moves. If a landlord knows the recommendation it receives is computed from rivals' non-public numbers, and knows those rivals are receiving recommendations computed partly from its own, the "we never talked" argument is doing less work than it appears to.

The second-order consequence matters more than the ruling itself. A revived claim is not a finding of liability — the Third Circuit allowed the case to proceed, which is a very different thing from deciding it. But survival past dismissal changes the economics of the dispute completely. Discovery in a multi-defendant antitrust case is where the real cost lives, and it lands on every named spoke, not just the hub. A mid-sized operator that licensed a pricing tool for a few thousand dollars a month can end up funding a document review that dwarfs the entire contract value. That asymmetry — small subscription, large downside — is the actual news for software buyers, and it is almost never priced into a vendor comparison.

This is not an isolated flashpoint either. Per the research underlying this analysis, the Department of Justice and state attorneys general have filed lawsuits against RealPage and other rental pricing software providers, and the FTC increased its enforcement focus on algorithmic collusion and pricing practices across 2024–2025. Private plaintiffs and public enforcers are now working the same theory from two directions. When civil and regulatory pressure converge on one product category, vendors historically respond by changing the product — which is a roadmap risk, not just a legal one.

Which Configuration Wins Which Job

Forget vendor names for a moment and compare three architectures a pricing tool can use. This is the side-by-side that no single source article offers, because each source is covering the case rather than the buying decision.

Architecture 1 — internal-data-only. The model sees your occupancy, your renewals, your history. Wins Job A cleanly. Antitrust exposure on the hub-and-spoke theory: essentially nil, because there is no shared competitive input to pool. Weakness: it is blind to a sudden market shift until your own conversion rates reveal it, which is a lagging signal.

Architecture 2 — public-data-augmented. Internal data plus scraped listing prices anyone could look up. Wins Job A plus a weaker version of Job B. Exposure: substantially lower, since the input is information already available to the whole market. Weakness: public asking rents are noisy and often stale relative to what leases actually sign at.

Architecture 3 — pooled non-public competitive data. Rivals contribute confidential executed-lease data; the model returns a recommendation informed by all of it. Strongest at Job B. This is the configuration that Cornish-Adebiyi puts on the table, and the one where accepting recommendations at a high rate looks, to a plaintiff, like adherence to a scheme.

Run the comparison honestly and Architecture 3 wins on raw pricing precision. That is why it sells. The question a buyer should ask is whether the incremental precision over Architecture 2 justifies sitting inside a diagram where a court might place you as a spoke. Our read: for most operators below institutional scale, it does not, because the marginal revenue lift from pooled data is a percentage-point conversation while the litigation downside is a fixed-cost conversation, and fixed costs do not scale down with your portfolio.

The AI Wrinkle That Makes This Harder, Not Easier

Machine-learning pricing models sharpen the problem in a way that older rules-based software did not. A model can learn that certain pricing patterns produce better outcomes and converge on parallel behavior across all its customers without any human ever agreeing to anything — the coordination becomes an emergent property of shared training data rather than an act. Traditional antitrust frameworks were designed around human conspiracy, and they strain here.

The uncomfortable implication for AI-driven workflow automation generally: the more autonomous the tool, the harder it is for a defendant to point at a human decision and call it independent. "The model decided" is not obviously a defense. It may be the opposite. This is the same governance gap Cyber Sentinel Daily documented in AI deployments lacking access controls — organizations adopting automated systems faster than they build the oversight to explain what those systems did and why.

The Switching Cost Nobody Prices In

If a team concludes it needs to move off pooled-data pricing, the migration pain is real and worth naming before signing anything.

The data export reality is the first wall. Pricing tools accumulate years of model-tuned history, and what exports is usually the raw inputs, not the fitted model. A team that has run three years of algorithmic pricing may have let its own internal pricing judgment atrophy — there is no institutional memory to fall back on, because the software was the memory. Rebuilding a manual or internal-only pricing process is a staffing question, not a software question.

The second wall is contractual. Data-contribution clauses, retention terms, and whether you can compel deletion of your submitted lease data are exactly the provisions nobody negotiates at signing and everybody wants at exit. Check them before the renewal, not after.

Bottom Line

The bottom line: Cornish-Adebiyi did not decide that algorithmic pricing is illegal, and any commentary claiming it did is overreading a procedural ruling. What it did was make the shared-competitive-data architecture expensive to defend, and on balance the more likely outcome is not a wave of judgments but a quiet product shift — vendors decoupling the internal-optimization engine from the competitive-data feed so buyers can purchase Job A without inheriting Job B's exposure. Operators who ask for that separation now will get it sooner than those who wait for a verdict.

For teams evaluating business tools of any kind, the transferable lesson is narrower and more useful than the headline: when a product's core value comes from pooling data across competitors, the data-sharing terms are not a legal footnote. They are the product.

Frequently Asked Questions

Disclaimer: This article is editorial commentary for informational purposes only and is not legal advice. It does not reflect independent product testing. Software features, pricing, and contract terms change; verify current details with the vendor and consult qualified counsel on antitrust matters. Research based on publicly available sources current as of August 4, 2026.