US algorithmic accountability

A Narrower Rent-Algorithm Law Would Survive the First Amendment That New York's Section 3 Did Not

A federal court blocked New York's ban on following rent-software recommendations because it ignored public versus non-public data. DOJ's RealPage decree shows the fix.

New York's Rent-Algorithm Law After the Injunction People of Internet Research · US 12 months Training data age floor DOJ decree limits model training t… Central Hudson Scrutiny standard applied Court treated recommendations as c… Sep 29 Injunction date SDNY preliminarily enjoined Sectio… peopleofinternet.com
New York's Rent-Algorithm Law After th… People of Internet Research · US 12 months Training data age floor Central Hudson Scrutiny standard applied Sep 29 Injunction date peopleofinternet.com

Key Takeaways

The strongest case for New York's law is simple. If competing landlords feed confidential lease data into one vendor's model, and that model returns rents they then follow, the result can look like a cartel with a software front end. Courts and enforcers have treated that as a real antitrust harm. A state that wants to stop it before it spreads has a legitimate interest, and the 2025 statute was a serious attempt.

On September 29, 2026, Judge Valerie Caproni of the Southern District of New York preliminarily enjoined one part of it. The challenge is RealPage, Inc. v. James. As Vinson & Elkins reports, the court blocked Section 3 and left Section 2 in force. The Attorney General then asked the court to reconsider, according to the same account. The ruling is preliminary, and an appeal remains possible. But the reasoning is worth taking seriously, because it describes a way to regulate algorithmic pricing that holds up.

What the statute does, and which part fell

General Business Law § 340-b has two operative prohibitions. As the statutory text on the New York Senate site shows, subdivision 2 makes it unlawful to "knowingly or with reckless disregard facilitate an agreement" among owners or managers "to not compete," including by operating or licensing software or an "algorithmic device" that performs a coordinating function. Subdivision 3 reaches the landlord. An owner or manager who knowingly or recklessly sets or adjusts rents, renewal terms, or occupancy levels based on such recommendations is deemed to have entered an unlawful agreement.

The court enjoined that second prohibition, the one the opinion calls Section 3. It held that algorithmically generated pricing recommendations are commercial speech and applied intermediate scrutiny under Central Hudson, not strict scrutiny. That is a modest ruling for the state. The law still failed, because Section 3 was, in the court's words as quoted by Vinson & Elkins, "undiscerning in targeting software without regard to its use of public versus non-public data."

The judge also named the alternative that would have been narrower. The legislature could have limited the prohibition to software performing a coordinating function "based on non-public data from horizontal competitors." Section 2, by contrast, survived at this stage because it does not make new conduct unlawful. It sits on top of existing antitrust law. The state's motion to dismiss was denied.

Why the public/non-public line matters

A rent recommendation built from public listings, published market indices and a landlord's own history is a form of analysis. Think of a broker's comparable-rents memo or a subscription market report. Banning a landlord from acting on it restricts speech and ordinary business judgment, and it does so even where no competitor's secrets are pooled. Section 3 swept in all of that because it turned on the presence of software, not on the presence of coordination.

The antitrust concern is different. It arises when a vendor takes nonpublic, competitively sensitive inputs from rivals and returns outputs that align their prices. That is the conduct the federal government has targeted. The Justice Department's November 24, 2025 proposed settlement with RealPage would, if approved, require RealPage to stop using competitors' nonpublic, competitively sensitive information to set rents in live operation. It would also limit model training to nonpublic data aged at least 12 months, and bar models whose geographic effects are narrower than a state. It would remove features that limited price decreases and require a court-appointed monitor.

That decree draws the line Judge Caproni found missing in New York's statute. It does not prohibit rent analytics. It prohibits pooling rivals' confidential data and using it in real time. The remedy is aimed at the mechanism of harm.

The costs of the broader approach

Overbroad algorithm bans carry three costs.

What lawmakers should take from this

The pro-tenant goal and the pro-innovation goal are not in conflict here. A tightly drawn rule can serve both. Four design choices follow from the ruling and the federal decree.

The Attorney General's reconsideration motion may or may not succeed, and the Second Circuit may eventually weigh in. Either way, the lesson is already visible. Algorithmic accountability works best when it targets what makes an algorithm anticompetitive, namely shared secrets and aligned prices, rather than the fact that it is an algorithm. New York can still pursue that goal. It should amend Section 3 along the lines the court itself sketched, rather than defend a version that treats public-data analysis as price-fixing.

Sources & Citations

  1. N.Y. Gen. Bus. Law § 340-b (NY Senate)
  2. DOJ: RealPage proposed settlement press release (Nov. 24, 2025)
  3. Vinson & Elkins: court enjoins Section 3
  4. Multifamily Dive: RealPage injunction coverage