plainstamp

HUD FHEO AI tenant screening builder's guide

Rule: us-hud-fheo-ai-tenant-screening-2024. Source: HUD/OFHEO press release, May 2, 2024 — links to two PDF guidance documents. Statutory framework: Fair Housing Act, 42 U.S.C. §§ 3601-3631; HUD disparate-impact rule, 24 CFR § 100.500. Audience: prop-tech engineering teams, property-management software vendors, property-management operators using AI / algorithmic tenant-screening tools. Severity: mandatory (the underlying Fair Housing Act applies; the May 2024 guidance clarifies how it applies to AI / algorithmic decisions).

What HUD did, in plain English

On May 2, 2024 the U.S. Department of Housing and Urban Development released two guidance documents through its Office of Fair Housing and Equal Opportunity (FHEO):

  1. Guidance on Application of the Fair Housing Act to the Screening of Applicants for Rental Housing — addresses tenant-screening AI / algorithmic systems used to predict tenancy success, evaluate criminal-record histories, eviction-record histories, and credit-screening data.
  2. Guidance on Application of the Fair Housing Act to the Advertising of Housing, Credit, and Other Real Estate-Related Transactions through Digital Platforms — addresses targeted-advertising AI used by digital platforms in housing-related transactions.

Together they reaffirm a single bottom-line rule: the Fair Housing Act's disparate-impact framework applies to algorithmic and AI-based decisions exactly as it applies to human decisions. A tenant-screening tool that produces a disparate impact on a protected class (race, color, national origin, religion, sex, disability, familial status) is unlawful unless the housing provider can show (a) the tool is necessary to achieve a substantial, legitimate, nondiscriminatory interest AND (b) no less-discriminatory alternative is available.

This is not a new statute. It is HUD's official position on how the existing Fair Housing Act applies to a fast-growing category of automated housing decisions.

Who this applies to

The guidance reaches:

If you are only a software vendor and your customers are housing providers, you are not the directly-liable party under FHA — but your customers will demand contractual representations and audit cooperation. Building disclosure and dispute-process scaffolding into the tool is now table stakes.

The five required elements

The plainstamp us-hud-fheo-ai-tenant-screening-2024 rule encodes five required elements, each derived from the May 2024 guidance.

1. Individualized assessment

Adverse decisions (denial, conditional approval, increased deposit, restricted unit options) must rest on an individualized assessment of the applicant. Algorithmic outputs are advisory inputs only.

This is the most-violated element in the field. A tenant-screening tool that returns a "decline" recommendation and is auto-actioned without a human or process review is non-compliant. Even if the human review is procedural, it must consider mitigating circumstances, recency and relevance of any criminal or eviction record flagged, and disability-related context.

2. Disclosure of data sources and prediction targets

Before or at the time of an adverse action, the applicant must receive a disclosure naming the screening tool, identifying the data categories the tool consults, and identifying the specific prediction targets the tool generates. Generic "automated decisioning" language does not satisfy this element.

In practice this means: the adverse-action notice names the vendor (e.g., "SafeRent screening report"), the data categories ("credit history, eviction filings within prior 7 years, criminal records within prior 7 years, income verification"), and the prediction targets ("predicted timely-rent-payment likelihood, predicted lease-violation risk").

3. Dispute and correction process

The applicant must be informed in writing of (a) the right to dispute any data input the tool used, (b) the procedure for requesting correction of inaccurate records, and (c) a reasonable period during which the applicant may submit corrections, mitigating evidence, or a request for individualized review before the decision becomes final.

For tenant-screening reports that meet the FCRA "consumer report" definition, the FCRA's adverse-action notice requirements (15 U.S.C. § 1681m) and dispute rights (15 U.S.C. § 1681i) apply on top — and the FCRA timelines are operationally good defaults: a 30-day dispute window with mandatory reinvestigation by the consumer-reporting agency.

4. Less-discriminatory-alternative monitoring

The housing provider must be able to demonstrate, in the disparate-impact analysis, that the tool is necessary AND that no less-discriminatory alternative would serve the same legitimate interest.

This is an ongoing obligation. As alternatives emerge in the market — narrower lookback windows on criminal records, individualized-assessment-first workflows, alternative scoring inputs that reduce protected-class disparity — the provider must reassess. Vendor-supplied claims of "bias-tested" or "fair" do not, by themselves, satisfy this element.

A practical implementation: an annual review (documented and retained) of the screening criteria, the tool configuration, and the available market alternatives. When the housing provider's portfolio includes a meaningful share of protected-class applicants, periodic disparity testing on the provider's own population is the cleanest evidentiary record.

5. No vendor delegation

Fair Housing Act compliance cannot be delegated to a third-party tenant-screening or AI vendor. The housing provider remains the responsible party for any disparate-impact outcome, regardless of contractual indemnification.

This mirrors the FINRA RN 24-09 doctrine in financial services and the EEOC technical-assistance position in employment AI: the provider is on the hook, period. Vendor representations are useful evidence but are not a defense.

Common failure patterns

After ~20 months of HUD enforcement and private-FHA litigation since the May 2024 guidance, four failure patterns recur:

  1. Auto-action on tool output. A screening tool returns "decline" and the property-management workflow auto-issues the adverse-action letter without any human or individualized-assessment step. The HUD position is unambiguous: this is a per-se violation regardless of whether the underlying decision would have been the same after individualized review.
  2. Generic "automated decisioning" notice. The adverse-action notice says only that "an automated screening tool was used." Without naming the tool, the data categories, and the prediction targets, the applicant cannot meaningfully dispute. This is the most-cited disclosure failure in the early enforcement actions.
  3. Vendor-said-it-was-fair defense. Provider relies on a vendor representation that the tool is "bias-tested" or has been audited for "fairness." The FHA disparate-impact framework asks whether the tool produces a disparate impact in the provider's specific deployment context — which the vendor's audit cannot answer.
  4. Long-lookback criminal-record screening. A tool that flags any criminal record within a 10-year window (or longer) is increasingly indefensible under the less-discriminatory-alternative prong: HUD has signaled that 5-year (or shorter) windows for non-violent records are widely available and equally effective for legitimate interests. Providers using long lookbacks should expect to defend the choice.

Stacking with adjacent regimes

The HUD FHEO guidance does not stand alone. Builders implementing AI tenant-screening will, in practice, also be subject to:

Minimum-viable-compliance checklist

For a housing provider operating a small-to-mid-sized portfolio that uses a third-party tenant-screening tool with AI-derived scoring:

Sample disclosures

Plain-language adverse-action notice (≤200 words)

Your rental application was reviewed using [tenant-screening tool name], a third-party screening report. The tool considered: credit history, eviction filings within the prior 7 years, criminal records within the prior 7 years, and income verification. Algorithmic outputs were one input; the final decision was reviewed individually by [property manager] who considered all of your application materials.

If any data the tool used is inaccurate, or if you have additional context — for example, evidence of rehabilitation, a record sealing or expungement, or a request for reasonable accommodation — you have 30 days from the date of this notice to dispute or submit supplemental information. Contact [housing provider contact] in writing to dispute or supplement.

[Housing provider] is responsible for this decision under the Fair Housing Act and cannot delegate that responsibility to a screening vendor. If you believe this decision was discriminatory based on a protected class, you may file a complaint with HUD at hud.gov/fairhousing or with [state fair-housing agency].

Formal-language adverse-action notice (legal-counsel-grade)

NOTICE OF ADVERSE TENANT-SCREENING DECISION

Pursuant to the Fair Housing Act (42 U.S.C. §§ 3601 et seq.) and HUD Office of Fair Housing and Equal Opportunity guidance dated May 2, 2024 ("Guidance on Application of the Fair Housing Act to the Screening of Applicants for Rental Housing"), [housing-provider name] hereby discloses:

  1. Tool used: [vendor name and product identifier].
  2. Data sources consulted: [enumerated categories].
  3. Prediction targets generated: [enumerated].
  4. This decision rests on an individualized assessment of your application materials by [reviewer name and title]. Algorithmic outputs were advisory.
  5. You have the right under this notice to dispute any data input the screening tool used, to submit mitigating evidence (including evidence of rehabilitation, record sealing or expungement, or a reasonable accommodation request), or to request individualized review. Submissions made within 30 days of the date of this notice will be reviewed before this decision becomes final. Submissions to: [contact address].
  6. Where the screening report constitutes a "consumer report" under 15 U.S.C. § 1681a(d), you have additional FCRA rights including the right to obtain a free copy of the report from [consumer reporting agency] and to dispute inaccuracies under 15 U.S.C. § 1681i.
  7. [Housing provider] retains full responsibility for this decision under 42 U.S.C. § 3604 and cannot delegate that responsibility under any contract with [vendor].

What "compliant" looks like in code

If you are a tenant-screening software vendor or a property-management platform integrating a screening API, the operational implications are:

The plainstamp rule encodes these as the five required elements; lookup_disclosure({jurisdiction: "us", channel: "ai-generated-content", use_case: "housing"}) returns them in citation-grounded form. validate_disclosure does a heuristic check that a candidate disclosure mentions the required elements — useful as a pre-deploy sanity check, not a sufficiency determination.

Authoritative sources

Disclaimer

Not legal advice. plainstamp surfaces the published text of real regulations and the May 2024 HUD guidance, with citation back to HUD's published source. For high-stakes housing-provider deployments, verify against the cited HUD documents and consult counsel licensed in your jurisdiction. The disparate-impact analysis under 24 CFR § 100.500 is fact-intensive and the right answer depends on the provider's portfolio, the tool's configuration, and the available market alternatives.