The Algorithm as Legal Laundering
What RealPage does, what the DOJ actually stopped, and why the settlement leaves the machinery running
The Justice Department filed a Proposed Final Judgment on Wednesday in United States v. RealPage. It’s the antitrust case against the algorithmic revenue management software landlords have been using to coordinate rent prices across most of the American rental market.
Here’s what the algorithm did. Landlords entered their pricing intent into RealPage’s software. The software calculated a “recommended” rent that accounted for the pricing data other landlords had submitted. Landlords adopted the recommendation. Rents rose in unison. The DOJ alleged this violated Section 1 of the Sherman Act, because landlords using the platform effectively acted in concert on pricing rather than competing independently.
That is price-fixing. If landlords met in a hotel room to compare rent sheets, we would call it a cartel. Route it through an algorithm and the DOJ needed twenty months to file.
What the settlement actually stops.
Willow Bridge Property Company, a co-defendant, is enjoined from licensing or using revenue management software that relies on competitively sensitive data. Willow Bridge is prohibited from sharing competitively sensitive information with other landlords. Willow Bridge must establish an antitrust compliance policy. Willow Bridge must cooperate with the United States in this litigation.
Read that list again. It’s a settlement with one landlord about one landlord’s future behavior. RealPage itself, the software vendor whose product coordinated the market, is not in this Proposed Final Judgment. That’s coming in a separate track.
The remedy targets the participant, not the machinery. Willow Bridge stops. The revenue management infrastructure it used continues. Every other landlord licensing similar software continues.
The public comment problem.
The DOJ’s May 8 Response to Public Comments notes that the Antitrust Division received eight public comments on the Proposed Final Judgment. Eight.
This is a settlement that affects the rents of millions of American tenants. Eight comments. The Antitrust Procedures and Penalties Act invites public participation, and the public was not there.
I want to name what that means without moralizing. The DOJ ran the process it’s required to run. Tenants overwhelmingly did not know it was running. That gap between formal transparency and actual participation is not accidental. Federal Register comment periods are legible to lobbyists, to industry counsel, and to a small number of academic and civil society specialists. They are not legible to the people whose rent went up.
Meanwhile, other things landed the same day.
While the DOJ was filing on RealPage, NormSense’s corpus caught Zoomcar filing three separate SEC 10-K disclosures. All on July 16. All AI-related. All about algorithmic pricing and algorithmic determination:
Zoomcar’s 10-K/A disclosed algorithmic damage coverage and protection determination for guests without explaining contestability or error correction procedures.
Zoomcar disclosed algorithmically determined ratings-based listing positioning that governs host marketplace visibility, without disclosed ranking criteria.
Zoomcar disclosed algorithmic dynamic pricing and trip protection fee determination for guests, without transparent explanation of pricing logic.
Three separate deployments. Same company. Same day. All disclosed to the SEC because SEC disclosure is legally required. None of them disclosed to the guests, hosts, or drivers whose lives the algorithm actually shapes.
The NormSense corpus captures 39 observations across the SEC AI Risk Disclosure Requirements cluster. Seven orgs on record supporting it. Zero on record opposing it. Disclosure to the SEC is a mature norm defended by named regulators, corporate legal, and board governance interests. It’s stable.
Disclosure to affected persons is not that. The corpus has an emerging cluster on state-level AI transparency frameworks with 97 observations. It also has 63 observations of NYC employers deploying automated employment decision tools without completing required pre-deployment bias audits. Non-compliance at scale, with nobody on record defending the non-compliance and nobody effectively enforcing against it.
This is the pattern the corpus is catching. Disclosure to markets is mature and defended. Disclosure to people is emerging, contested, and under-enforced. The same design pattern that made RealPage’s rental coordination possible in the first place.
The broader field.
Two other norms in the NormSense corpus point at the same architecture. Online personalized pricing without meaningful consent, where retailers charge different customers different prices based on personal data without telling the customer their price is personalized. And Illinois retailers deploying surveillance pricing algorithms that use personal data to set individualized prices, again without disclosure to the customer.
Neither cluster carries direct organizational contestation yet. Nobody’s on record defending them on the merits. Nobody’s on record opposing them at scale either. They are the kind of practice that operates in the space between what’s technically disclosed and what’s actually known, and that space is where enforcement doesn’t reach.
Different sectors. Same mechanism. An algorithm makes a coordination or discrimination decision that would trigger legal scrutiny if humans made it explicitly. Route it through code, and the legal apparatus lags for years before catching up. By the time it does catch up, the pattern has moved. The remedy targets last generation’s plumbing. The next generation’s plumbing is already running.
That is what “algorithm as legal laundering” means in practice. It’s not that algorithms are unaccountable in principle. Section 1 of the Sherman Act applies. State consumer protection statutes apply. Common law fraud applies. The problem is that the legal apparatus is calibrated to a pace of coordination that assumes humans in rooms. Algorithms coordinate at a different pace, in a different medium, and the enforcement lag is where the real value extraction happens.
Watch what stays.
Willow Bridge will comply. RealPage’s next generation product will comply. Zoomcar will keep filing 10-Ks that satisfy the SEC and reveal nothing to the people who use its platform. Retailers will keep running personalized pricing that never appears in a price tag.
Eight people commented on the RealPage settlement.
The people who catch this stuff, in real time, at any meaningful scale, are the civil society researchers, the state AGs willing to invest in multi-year investigations, and platforms like NormSense that read the disclosures no one else reads.
That’s what we are trying to catch. Not who said what at what conference. What actually stays when the microphones turn off.
Zach, see you in the cluster pages.