More screening products now hand landlords a single score or a "recommended" result instead of raw reports. Some of those scores come from automated models, and some vendors market them as AI. For a small landlord, that can feel like a shortcut. It is also a new source of legal risk.
The rules did not disappear because a computer made the first cut. Fair housing law, the Fair Credit Reporting Act (FCRA) and a growing set of state rules all still apply to you, the person who makes the rental decision. This guide explains how each one reaches algorithmic screening, what changed between 2024 and 2026, and a practical checklist for keeping your process defensible.
Note: This article is general education, not legal advice. Fair housing, consumer reporting and automated decision rules vary by state and city and are changing quickly. Talk to a local attorney before you change your screening policy.
Why automated screening scores create fair housing risk
A screening score is only as fair as the data and weights behind it. A model does not need to consider race, national origin or disability to produce results that fall harder on a protected group. If it leans heavily on credit history, eviction filings or criminal records, it can carry existing disparities straight into your approval decisions.
That is the core of disparate impact. The Fair Housing Act bans intentional discrimination, and in 2015 the U.S. Supreme Court confirmed in Texas Department of Housing and Community Affairs v. Inclusive Communities Project that it also reaches policies with an unjustified discriminatory effect. A neutral-looking score cutoff can be one of those policies.
A real case shows how this plays out. In November 2024, a federal court in Massachusetts approved a settlement of more than $2.2 million in Louis v. SafeRent Solutions, according to plaintiffs' counsel Cohen Milstein.
The suit alleged that a screening score relied too heavily on credit data and ignored housing vouchers, which hurt Black and Hispanic voucher holders. As part of the settlement, the vendor agreed to stop including its score in certain cases, including for applicants using vouchers.
The lesson for a small landlord is simple. You can be the one answering a complaint even when a vendor built the model. Buying a score does not transfer your fair housing duties. Our fair housing compliance guide covers protected classes and reasonable accommodations if you want the broader picture.
What changed at HUD between 2024 and 2026
On May 2, 2024, HUD announced two guidance documents that addressed algorithms directly. One covered screening of applicants for rental housing, including screening that relies on algorithms and artificial intelligence. The other covered advertising through digital platforms. Per HUD's release, housing providers that use third-party screening companies, including those using AI, must comply with the Fair Housing Act.
The federal posture then shifted. According to a notice HUD published in the Federal Register on April 6, 2026, HUD withdrew a group of fair housing guidance documents effective September 17, 2025. That list included the April 29, 2024 digital advertising guidance and HUD's 2022 guidance on the use of criminal records. HUD states that withdrawn documents should not be relied on as authoritative.
The 2024 screening guidance was not on that withdrawal list, but you should not treat it as a firm federal position either. Read it as a useful description of risk, not a safe harbor.
HUD is also rewriting its disparate impact regulation. In January 2026, it proposed removing the burden-shifting framework in its Fair Housing Act rules and leaving the standard to the courts. An August 10, 2026 supplemental proposal set a comment deadline of October 9, 2026. As of this writing, those changes are proposed, not final.
What has not changed is the statute. HUD's withdrawal notice says actions that do not comply with the text of the Fair Housing Act remain subject to enforcement, and private parties can still go to court. Many state and local fair housing laws also have their own effect-based standards.
How the FCRA applies when a score drives your decision
Most screening scores are built from consumer reports. That brings the Fair Credit Reporting Act into play, no matter how sophisticated the model is.
Accuracy
Screening companies must follow reasonable procedures to keep reports accurate. In October 2023, the FTC and CFPB announced a $15 million settlement with TransUnion's rental screening unit. The agencies alleged it reported dismissed eviction cases as judgments, split single cases into multiple evictions and let sealed records appear. Our article on how accurate tenant screening reports are and where errors come from goes deeper.
Adverse action notices
Per the FTC, if anything in a consumer report contributes to a denial, a cosigner requirement or a higher deposit, you must give the applicant an adverse action notice. That holds even when the report played only a small part.
The notice names the screening company and its contact details. It also states that the company did not make the decision and explains the right to a free report within 60 days and the right to dispute.
If you used a credit score, the FTC says the notice must also include the score, its range and the key factors that hurt it. A vague "did not meet our score" line is not enough. Point applicants to our guide on how to dispute a tenant screening report step by step if they believe the data is wrong.
Disposal
The FTC also requires you to securely dispose of consumer reports and information taken from them when you are done using them. Shred paper. Delete electronic files so they cannot be reconstructed.
State and local rules on automated screening
State and city rules are where AI-specific disclosure requirements are actually appearing. The picture is uneven, so check your own jurisdiction.
| Jurisdiction | Rule | Status as of October 2026 |
|---|---|---|
| Colorado | SB 26-189, automated decision-making technology in consequential decisions, including residential real estate | Signed May 2026; scheduled to take effect January 1, 2027 |
| California | Privacy agency rules on automated decisionmaking technology for significant decisions, including housing | Adopted 2025; automated decision provisions apply from January 1, 2027, for businesses covered by the state privacy law |
| New York State | A3125-A and S6471-A, audits and notice for automated housing decision tools | Proposed; not law |
| New York State | A9294-A, written screening criteria before any application fee | Proposed; not law |
| New York City | Fair Chance for Housing Act, limits on criminal history screening | In effect since January 1, 2025 |
Colorado
Colorado replaced its 2024 AI Act with SB 26-189, which Governor Polis signed in May 2026. According to summaries from Lathrop GPM and Eckert Seamans, it requires notice before covered technology is used in consequential decisions, including residential real estate. If it materially contributes to an adverse decision, the deployer must explain its role within 30 days. Attorney General rulemaking is still pending.
California
California's Civil Rights Council adopted rules on automated-decision systems that took effect October 1, 2025, but those rules address employment. For housing, the more relevant rules come from the California Privacy Protection Agency. They cover automated decisions about housing, with compliance beginning January 1, 2027. They apply to businesses that meet state privacy law thresholds, which often means your screening vendor rather than you.
New York
New York has two pending bills worth watching. A3125-A and its Senate companion S6471-A would require landlords who use automated housing decision tools to obtain annual independent disparate impact audits, publish a summary and notify applicants in advance. A9294-A would require written screening criteria before collecting any application fee. Neither is law as of this writing.
New York City already requires a two-step process under its Fair Chance for Housing Act. Landlords must conditionally approve an applicant before reviewing criminal history and then conduct an individualized assessment before denying.
A practical checklist for landlords using automated screening
You do not need a compliance department to screen responsibly. You need a written process that you follow every time. Our tenant screening compliance requirements playbook walks through each step in more depth.
- Ask vendors how the score is built. Ask which data sources feed it, how each factor is weighted, whether it has been tested for disparate impact and whether you can see the underlying reports.
- Keep a human decision. Treat any score or recommendation as one input. You should review the reports and make the final call yourself.
- Write your criteria before you advertise. Set minimum income, credit and rental history standards in advance and apply them to every applicant for the unit.
- Use individualized assessment. Look at the age and relevance of any criminal or eviction record, and consider context and mitigating information.
- Count all lawful income. A score that ignores vouchers or other assistance can create problems where source-of-income laws apply.
- Send complete adverse action notices. Include every FCRA element and any credit score disclosure, every time a report contributes to an unfavorable decision.
- Keep records. Save your criteria, the reports you reviewed, your reasoning and copies of notices, and dispose of report data securely when done.
Source of income protections are expanding. See our overview of source-of-income discrimination laws and what landlords need to know before you rely on a score that does not account for vouchers or other rental assistance.
Frequently asked questions
Is it illegal for landlords to use AI tenant screening tools?
No federal law bans them. However, the Fair Housing Act and the FCRA still apply, and some states, such as Colorado, have adopted notice rules for automated decisions.
Can a landlord blame the screening company for a discriminatory score?
Not reliably. The landlord makes the rental decision, so the landlord can face fair housing claims even when a vendor designed the scoring model.
Did HUD cancel its AI tenant screening guidance?
HUD withdrew its 2024 digital advertising guidance and its 2022 criminal records guidance effective September 17, 2025. The 2024 screening guidance was not on that list, and the Fair Housing Act itself still applies either way.
Do landlords need to send an adverse action notice when a score causes a denial?
Yes, if the score or any report behind it came from a consumer report. Per the FTC, the notice must name the screening company and explain the applicant's rights.
What to do next
Automated scores make screening faster, but they also make it easier to lose track of why an applicant was denied. The landlords who stay out of trouble write their criteria down, review the full reports, make the decision themselves and tell renters why.
Shuk supports that kind of process. Landlords set their own pass criteria, such as minimum credit score, rent-to-income ratio and policies on evictions and bankruptcies, before screening begins. Credit, criminal, eviction and other screening reports run through a screening partner and are charged to the applicant, with optional income and employment verification.
Landlords can request documents such as proof of income, and Shuk never auto-approves or denies an applicant. When a landlord declines, renters see the decision and the reasons the landlord selected.
At as low as $2 per unit per month, with no setup fees and no contract, and with White Glove Onboarding included at no additional cost, Shuk makes consistent, human-led tenant screening feasible for landlords and property managers running 1 to 100 units.
Book a demo at shukrentals.com/book-a-demo to see how landlord-set pass criteria, partner screening reports and document requests work together so every decision stays yours and stays documented. For more guides like this one, browse our tenant screening hub with guides on reports, criteria and compliance.
