The RealPage case was about rent prices. Its lessons reach every brokerage that uses AI.

In August 2024 the U.S. Department of Justice and eight states sued RealPage, the maker of widely used rent-setting software. In January 2025 the DOJ added six large landlords as defendants. In May 2026 a federal court approved the DOJ's settlement with RealPage. The states' claims and the claims against some landlords are still going.

Most real estate brokers read the headlines, filed the case under "property management problem," and moved on. That is understandable, but it misses something.

What the case was actually about

It is worth being precise, because the case is easy to overstate. RealPage is an antitrust case. The government alleged two things:

• Landlords shared nonpublic, competitively sensitive rent and lease data through RealPage's software. That pooled data then shaped the rents the software recommended back to competing landlords.
• RealPage unlawfully maintained a monopoly in the market for that kind of software.

The settlement reflects that focus. RealPage may no longer use competitors' nonpublic data to set rents in real time. It may train its models only on nonpublic data that is at least twelve months old. It also had to remove or redesign features that pushed landlords to accept its recommendations, and a court-appointed monitor will oversee compliance. When the settlement was announced, the head of the DOJ's Antitrust Division put the principle plainly: Competing companies must make independent pricing decisions, and that remains true when an algorithm is involved.

So RealPage did not create a general rule that a business is liable for everything a licensed algorithm does. It applied old antitrust law to a new tool.

Why it still matters to a brokerage

Three things about the case do carry over to any firm that uses AI tools.

First, the businesses that used the tool were named, not only the vendor. Six landlords became defendants, and several have since settled with the government. "We just used the software" did not keep them out of the case.

Second, regulators are applying existing law to AI tools without waiting for new AI statutes. In 2023 the Federal Trade Commission barred Rite Aid from using facial recognition for five years, saying the company failed to vet its vendors or test and monitor the technology it deployed. In 2025 the Massachusetts attorney general settled with a lender over the disparate effects of its AI underwriting models. Neither case needed an AI law.

Third, new laws are arriving anyway. California and New York both passed laws in 2025 aimed at algorithmic pricing, and more than a dozen cities and counties have restricted algorithmic rent-setting. Tenants have already filed suits under several of them.

A pattern brokers have seen before

Website accessibility is a useful comparison. Federal lawsuits under Title III of the Americans with Disabilities Act rose from 2,722 in 2013 to a peak of 11,452 in 2021, and there were still 8,667 in 2025. A large share of that growth came from a small number of plaintiff firms. In 2024 a single firm filed 2,598 federal ADA cases in California alone.

Few businesses saw it coming. Brokerages built websites for years without treating accessibility as a legal issue, because a website felt like marketing, not a public accommodation. Then the claims arrived in volume.

The comparison has limits, and they matter. AI claims will usually be harder to bring, because a plaintiff will need to show real harm, not just a barrier on a web page. On the other hand, where those claims succeed, actual damages will be in play, not just fees and a fix. That is NIRAP's reading of the trend, not a prediction anyone can guarantee.

There is one more difference. For private businesses, no regulation sets a technical accessibility standard, but courts and settlements settled on WCAG 2.1 Level AA as a practical benchmark. AI has general frameworks, such as the NIST AI Risk Management Framework and ISO/IEC 42001, but they were written mainly for organizations that build or operate AI systems at scale. There is not yet a widely used benchmark sized for a brokerage. A firm that adopts a clear standard early can point to it. A firm that waits may be measured against one that someone else chose.

Four exposures every brokerage now carries

1. Fair housing liability from AI-written content

AI tools write listing descriptions every day, and most of that copy is fine. Some of it is not. A phrase like "perfect for young professionals" or "ideal for empty nesters" can signal a preference based on familial status. The tool does not know it is doing that. The agent who publishes without reading carefully is the one who answers for it.

2. Misrepresentation from AI errors

When a tool states a fact about a property, such as square footage, school boundaries, HOA fees or permitted use, and the agent passes it to a client without checking, the agent has made that representation. If the tool was wrong, "the AI told me" is not a defense.

3. Screening decisions made with automated tools

Brokerages that screen rental applicants, directly or through a property management arm, are responsible for the outcome under the Fair Housing Act whether a person or a product made the call. HUD's 2024 guidance on screening tools has been removed from its website, but the statute did not change. Vendors face liability too: in 2024 the screening company SafeRent agreed to a settlement of about $2.3 million in a case alleging its scoring harmed Black and Hispanic applicants who used housing vouchers.

4. Antitrust exposure from pricing tools

This is the RealPage lesson in its narrow form. Any tool that recommends prices, rents or terms using competitors' nonpublic data deserves a hard look, especially for brokerages with property management operations. California's new law covering pricing algorithms is not limited to housing.

What documentation does, and what it does not

For each of these exposures, the most useful thing a brokerage can have is a record: a written AI policy, proof that agents were trained, a log showing that a person reviewed AI-assisted work, and notes on the risks the firm identified and what it did about them.

That record is evidence. It shows a regulator, a carrier or a court that the firm took reasonable steps, and it gives the firm something to stand on besides memory. It is not a safe harbor. No statute gives a brokerage reduced liability because it documented its AI governance, and no certificate, including NIRAP's, guarantees compliance. A firm with a record still has to defend itself. It simply has something to defend with.

Three questions every broker should be able to answer

1. Which AI tools are agents and staff using, who is using them, and for what?
2. What practices are in place to manage the risks those tools create?
3. If a regulator, carrier or opposing counsel asked for documentation tomorrow, what would you hand them?

These are not trick questions. They are the questions a license regulator could ask after a complaint, and the ones opposing counsel will ask after a deal goes wrong. Insurers have started adding AI exclusions to some liability policies, so carriers may ask them too.

What to do this quarter

• Take inventory. Ask agents and staff which AI tools they use. The answers will surprise you.
• Adopt a written policy. It does not need to be perfect. It needs to exist. NAR's template is a good start.
• Train your people. Most agents have never been told what a hallucination is or when AI output can be trusted.
• Keep records. Build the habit of noting when AI was used on client work and what a person checked.
• Have it checked by someone outside the firm. An independent review finds gaps while they are cheap to fix, and it gives clients and recruits something they can verify.