The new compliance lesson is blunt: an algorithm can assist a landlord, but it cannot become the landlord’s excuse.
Why Algorithms Became So Attractive
Corporate landlords operate at scale. A single regional property manager may oversee thousands of units across several states. Leasing teams need fast decisions, consistent standards, fraud controls, rent-payment predictions, income verification, identity checks, eviction history review, and criminal history screening. Screening vendors stepped into that pressure with automated reports and risk scores.
The sales pitch was powerful: fewer subjective decisions, less staff time, cleaner approvals, fewer risky tenants, and less human bias. But automation does not automatically remove bias. If the data feeding the model reflects unequal access to credit, unequal policing, uneven court records, source-of-income discrimination, unstable medical debt, or historic housing exclusion, the output may reproduce those inequalities while looking neutral.
The Fair Housing Problem
The Fair Housing Act prohibits housing discrimination based on protected characteristics such as race, color, religion, sex, disability, familial status, and national origin. Tenant screening can violate that law if it intentionally treats protected applicants differently or if a policy has an unjustified discriminatory effect and is not necessary to achieve a substantial, legitimate, nondiscriminatory interest.
Algorithmic screening creates risk because the landlord may not know how the tool weighs each factor. A score may be influenced by credit history, debt, eviction filings, thin credit files, address history, criminal records, income patterns, or other variables that correlate with protected status. If the landlord cannot explain why the decision is valid, relevant, and less discriminatory alternatives were considered, the landlord may be exposed.
Why The Guidance Rollback Did Not Make Algorithms Safe
HUD’s withdrawal of several FHEO guidance documents has encouraged some providers to think enforcement pressure is fading. That is only partly true. A withdrawn guidance document may no longer be authoritative HUD policy, but lawsuits do not depend only on HUD guidance. Courts, private plaintiffs, fair housing groups, state civil rights agencies, local ordinances, and consumer protection statutes can still matter.
This is why “HUD’s war” should be understood as broader regulatory and litigation pressure, not a single active memo. The war is against blind reliance on black-box screening. Even if federal agency priorities shift, a denied applicant can still ask a basic question in court: why did this system reject me, and did that method unlawfully screen out protected households?
If the landlord cannot explain the decision without hiding behind the vendor, the landlord has already lost control of the compliance file.
The SafeRent Lesson
The SafeRent litigation became a warning sign for the entire industry. Plaintiffs alleged that an algorithmic tenant screening score harmed Black and Hispanic voucher applicants by relying on factors such as credit history and non-rental-related debt while failing to account for voucher rent support and other relevant rental history. The case showed how a score that looks neutral can become the center of a discrimination claim.
The practical lesson is not that every screening score is illegal. The lesson is that a score must be defensible. If the tool penalizes applicants for factors that do not reliably predict tenancy performance, ignores stronger evidence of ability to comply with the lease, or produces worse outcomes for protected groups, landlords should expect scrutiny.
Back To Basics Means Relevant Criteria
The first basic principle is relevance. A screening policy should ask whether the applicant is likely to pay the tenant portion of rent, follow lawful lease terms, and avoid conduct that threatens resident safety or property operations. Criteria that do not serve those goals should be removed or narrowed.
That means a landlord should question broad credit-score cutoffs, automatic denials for old eviction filings, nonconviction criminal records, medical debt, student debt, thin credit files, or minor civil judgments. A person’s credit profile may say something about access to credit markets, but it may say very little about whether rent will be paid when a voucher covers most of the contract rent or when the applicant has a strong rental history.
Back To Basics Means Individual Review
Automated screening is most dangerous when it becomes final. A score can be useful as a flag, but it should not be the last word for close cases, voucher applicants, disability-related issues, VAWA-related adverse history, thin-file applicants, or applicants with records that may be inaccurate or stale.
Corporate landlords should build a human review layer. That review should allow applicants to submit corrected records, landlord references, proof of rent payment, proof of income, voucher documentation, evidence of rehabilitation, disability-related accommodation requests, domestic violence survivor explanations, and other information that the algorithm may not understand.
Back To Basics Means Written Standards
A screening policy should be written clearly enough that staff, applicants, vendors, auditors, and courts can understand it. The policy should say which factors are considered, which factors are excluded, what lookback periods apply, how voucher income is treated, when exceptions are available, how reasonable accommodation requests are handled, and how an applicant can appeal or correct information.
Vague standards invite inconsistency. Inconsistency invites fair housing claims. If one applicant is allowed to explain an old eviction but another is denied instantly, the landlord needs a reason that is not tied to race, disability, family status, national origin, voucher use, or another protected trait under federal, state, or local law.
Back To Basics Means Vendor Control
Corporate landlords often outsource screening and then treat the vendor’s report as objective truth. That is risky. A vendor may gather incomplete court records, misclassify cases, fail to distinguish filings from judgments, overstate criminal history, use outdated data, or apply a scoring model that the landlord does not understand.
Owners should require vendors to disclose what data is used, how errors are corrected, whether protected-class proxies are excluded, whether the model has been validated, how frequently the model is tested, what adverse action notices are generated, and whether the system can be audited. A landlord should not buy a black box and then claim surprise when the box produces discriminatory results.
Voucher Applicants Need Special Care
Voucher households expose one of the clearest weaknesses in algorithmic screening. A model that treats low income, weak credit, or limited assets as high risk may ignore the fact that a housing authority will pay a large share of rent directly to the landlord. If the applicant’s actual tenant portion is affordable, a full-rent income multiplier or credit-heavy score may be misleading.
In states and cities with source-of-income protections, this can become a legal problem quickly. Landlords may need to calculate income requirements based on the tenant-paid portion, process voucher paperwork, avoid “no Section 8” scripts, and evaluate applicants without penalizing them for using lawful rental assistance. The algorithm must fit the law, not the other way around.
Disability Accommodations Cannot Be Automated Away
Screening tools may flag poor credit, rental debt, criminal records, gaps in rental history, or prior lease problems. Some of those issues may be connected to disability. A tenant may have medical debt, a hospitalization-related rent disruption, a past behavioral incident now managed by treatment, or a support need that changes the risk analysis.
The Fair Housing Act requires reasonable accommodations when necessary for a person with a disability to have equal opportunity to use and enjoy housing. An automated denial process must pause when an accommodation request is made or when staff have enough information to recognize that disability-related review may be required. A landlord cannot let software overrule accommodation law.
Eviction Records Are Often Noisy Data
Eviction data can be misleading. A filing may not mean the tenant was evicted. The case may have been dismissed, sealed, resolved, filed during a pandemic disruption, or tied to conditions beyond the applicant’s control. Some jurisdictions have stronger tenant protections or sealing rules than others, making national screening harder.
A basic, defensible policy should distinguish filings from judgments, recent cases from old ones, nonpayment from serious lease violations, and unresolved records from corrected records. A landlord that denies every applicant with any eviction-related hit may look efficient, but efficiency is not a defense if the policy is inaccurate, overbroad, or discriminatory.
Criminal Records Require Narrow Tailoring
Criminal background screening remains one of the highest-risk areas. Arrests are not convictions. Old records may have little connection to present tenancy risk. Minor offenses may not predict danger to residents or property. Blanket bans can create discriminatory effects and may violate state or local law.
A back-to-basics approach asks whether the record is recent, serious, accurate, relevant to resident safety or property operations, and supported by an individualized review process. The policy should also allow applicants to dispute inaccurate records and provide mitigating evidence. Corporate landlords should not let a vendor’s criminal history flag become an automatic rejection machine.
Adverse Action Notices Must Be Useful
When a landlord uses consumer report information to deny an application, require a higher deposit, or impose less favorable terms, adverse action notice duties may apply under consumer reporting laws. These notices should not be treated as empty paperwork. They are part of the applicant’s ability to understand and challenge the decision.
A useful notice identifies the reporting company, explains the applicant’s rights, and gives enough information for the applicant to correct errors. If the landlord’s system cannot explain why the applicant was denied, that is not only a customer service problem. It is a compliance problem.
Corporate Landlords Need An Audit Trail
The larger the portfolio, the more important the audit trail. Owners should retain screening criteria, vendor contracts, model validation materials, adverse action notices, appeal records, accommodation requests, override decisions, voucher calculations, staff training records, and periodic outcome reviews.
A landlord should also test whether denial rates differ by property, market, protected-class proxy, voucher status, disability accommodation, family status, or language access need where lawful data analysis is possible. The goal is not to collect protected-class data recklessly. The goal is to know whether the system is producing patterns the company cannot defend.
What Going Back To Basics Looks Like
Going back to basics does not mean abandoning technology. It means technology must support a lawful screening policy. The landlord should define the criteria first, then choose or configure the tool around those criteria. The vendor should not be allowed to decide the company’s fair housing risk appetite through default settings.
A stronger model uses transparent criteria, shorter lookback periods, individualized review, voucher-aware income calculations, reasonable accommodation escalation, VAWA-sensitive review, meaningful appeals, accurate adverse action notices, and regular audits. That may be slower than instant scoring, but it is far safer than letting an unexplained number decide who gets a home.
Bottom Line
HUD’s pressure on tenant screening algorithms, even after later guidance rollbacks, has forced corporate landlords to confront a basic truth: automated screening is not a legal shield. The Fair Housing Act still applies, private lawsuits still matter, state and local protections still matter, and consumer reporting duties still matter.
The future of tenant screening is not blind faith in AI or a return to arbitrary human judgment. It is disciplined fundamentals: relevant criteria, transparent policies, human review, vendor accountability, accurate records, lawful voucher treatment, disability accommodations, clear adverse action notices, and documented appeals. The landlords that adapt will still use technology, but they will stop treating the algorithm as the decision-maker. In fair housing compliance, the basics are becoming the innovation.