Domain Atlas / Housing & homelessness services
SafeRent Tenant Screening Score
SafeRent Solutions (formerly CoreLogic Rental Property Solutions) sold landlords a Registry ScorePLUS tenant-screening model returning a single 200-800 'lease performance risk' score plus an accept/decline/conditional recommendation measured against a landlord-chosen cutoff (500 in the complaint's worked example, with an optional 450-499 conditional band), built from credit bureau reports and scores including non-tenancy debt, bankruptcy records, past-due accounts, payment performance, and eviction and landlord-tenant court records, weighted 'according to their statistical significance in predicting lease performance'; factor weights were undisclosed to landlords, applicants and the public, and the company's marketing told buyers a landlord cannot change the screening algorithm. The plaintiffs' pleaded theory concerns a missing input: as of 2021, HUD data cited in the complaint shows Black and Hispanic voucher holders in Massachusetts paying an average of $423/month toward rent and utilities while public housing authorities paid landlords an average of $1,159/month directly (at least 73.26% of the expected payment), on tenancies averaging over 21 years in the same unit, and that subsidy is not a model input. These are plaintiff-pleaded figures accepted by the court as allegations at the motion-to-dismiss stage only.[3]
What happened
SafeRent Solutions, formerly CoreLogic Rental Property Solutions, sells landlords a tenant-screening product whose Registry ScorePLUS model returns a single "lease performance risk" number between 200 and 800 plus an accept, decline or conditional recommendation. The inputs are credit bureau reports and scores including non-tenancy debt, bankruptcy records, past-due accounts, payment performance, and eviction and landlord-tenant court records, weighted "according to their statistical significance in predicting lease performance." Landlords choose the cutoff — 500 in the complaint's worked example, with an optional 450-499 "accept with conditions" band — and the vendor evaluates the score against those predetermined decision points and returns the answer. What landlords do not get is the weighting: factor weights were undisclosed to landlords, applicants and the public, and the company's own marketing told buyers that "a landlord cannot change the screening algorithm." One property using the tool described its practice in writing: "CoreLogic sends us a number, and if it is above the predetermined approved number, we move forward... We do not know why they were denied, other than their score was not high enough." Its leasing manager, the same record notes, "does not receive the detailed credit information at the time of running the applicant screening." When Mary Louis was denied at Granada Highlands in Malden on 27 May 2021 solely on her score, Metropolitan Management staff told her in writing: "we do not accept appeals and cannot override the outcome of the Tenant Screening."
The named plaintiffs were housing-voucher holders, and their pleaded theory was about a missing input rather than a faulty one. As of 2021, the complaint alleges citing HUD data, Black and Hispanic voucher holders in Massachusetts paid an average of $423 a month toward rent and utilities while public housing authorities paid landlords an average of $1,159 directly — at least 73.26% of the expected monthly payment government-guaranteed — and the average Massachusetts voucher holder stayed in the same unit for more than 21 years. None of that reaches the model; credit history does, and the complaint pleads its skew: median credit scores of 612 for Black consumers, 661 for Hispanic and 725 for White as of October 2021, with subprime shares of 45.1%, 31.5% and 18.3%, citing Urban Institute data. It further pleads 2019 Federal Reserve figures showing credit-derived risk reads rent risk through the wrong instrument: among consumers expecting to defer a bill, 45% would default on credit cards against 23% on rent or mortgage. Massachusetts had over 91,000 vouchers in use, roughly 24% of holders Black and 31% Hispanic, about 7% moving in a year, with over 1,500 Black and over 1,900 Hispanic voucher holders screened for an apartment annually — and a voucher holder who cannot lease up within an initial 120-day term must seek a discretionary extension or lose it, so each denial burns a hard search budget. Monica Douglas's denial at Millside at Heritage Park rested partly on an eviction record created when her landlord wanted the unit for a relative; her 22 July 2021 denial was reversed only after a personal appeal was rejected on 26 August and City Life/Vida Urbana, an outside tenant-advocacy group, drafted a second appeal delivered 1-8 September — roughly six weeks of organized advocacy for one reversal, worked through the landlord's channel rather than the vendor's. The Community Action Agency of Somerville, an organizational plaintiff, worked with 29 voucher tenants in the prior year, about 36% of its housing caseload, and reported diverting staff time for each credit-screen denial.
The class complaint was filed 25 May 2022. On 9 January 2023 the Justice Department and HUD filed a Statement of Interest arguing that tenant-screening companies are not exempt from the Fair Housing Act because decisions are made by algorithm; HUD General Counsel Damon Smith put it that "tenant screening policies are not exempt from the Fair Housing Act's protections just because decisions are made by algorithm." On 26 July 2023 Judge Angel Kelley denied the motions to dismiss the FHA §3604(a),(b) and Massachusetts c.151B race and source-of-income claims, reasoning that SafeRent "effectively controls" approval decisions because it alone builds and conceals the algorithm; the c.93A consumer-protection counts were dismissed. A memorandum of settlement followed on 22 December 2023, the agreement was executed and filed 28 March 2024 (Doc. 114/114-1), preliminary approval came 25 April 2024 and final approval on 20 November 2024: $2.275M total consideration, a $1.175M fund with no reversion, and service awards up to $10,000 each for two class representatives. The injunctive terms are the substance. For five years the company may return no SafeRent Score, no other tenant screening score, and no accept or decline recommendation on a report for a voucher-holder application, providing instead a report of underlying information for individualized landlord review (§3.5.2). For the "market" and "no-credit" products, landlords must affirmatively certify the applicant is not a voucher recipient before any score is returned; absent certification the score is suppressed by default (§3.5.3). Any tenant screening score may re-enter the voucher channel only if "found to be valid when used for voucher-holders by the National Fair Housing Alliance" or another organization mutually agreed by class counsel and the company (§3.5.5(ii)(1)) — though third-party credit scores such as the FICO score and VantageScore may still be passed through with disclosure of their source (§3.5.5(iii)). The company must train its housing-provider customers on the difference between its three models and explain why no score appears for voucher applicants (§3.5.4), and the court retains continuing and exclusive jurisdiction to enforce for five years from the company's certification of the changes (§6.9), placing the end of supervision around 2029-2030. How far those practice changes reach geographically is ambiguous on the record: the §3.5.2 and §3.5.3 text is not expressly limited to Massachusetts, while the classes, the notice population and the plaintiffs' framing are all Massachusetts-centered, and some coverage characterizes the changes as company-wide. SafeRent settled without admitting liability and stated it "continues to believe the SRS Scores comply with all applicable laws" (vendor claim). Mary Louis, on the system: "Everything is based on numbers. You don't get the individual empathy from them."
The sociotechnical reading
The court's motion-to-dismiss reasoning names the structural fact this case turns on, and it is an authority inversion rather than a model defect. Three parties can touch the decision and only one of them is the decision-maker on paper. The vendor, which has no housing relationship with anyone, authors the weights and returns the answer. Property management picks the cutoff, in consultation with the vendor, without knowing how the score is computed. The leasing desk that signs the lease receives a number, not the underlying credit detail, and wrote that it does not accept appeals and cannot override the outcome. Formal authority sat with the landlord; practical control sat with a company that never met the applicant, which is exactly why the court held that the screening company "effectively controls" approval decisions and could therefore be reached by the Fair Housing Act. The counterpart to that inversion is the missing input. The population being screened carries the strongest protective signal in the actual arrangement — a housing authority paying most of the rent directly, on tenancies averaging over 21 years — and that signal is not in the model's state at all, while a credit history the plaintiffs plead as a poor proxy for rent behaviour is. A risk model that cannot see the guarantee prices the guaranteed population as the risky one.
What makes this case unusual in the atlas is the remedy's location. Nothing in the settlement fixes the model, publishes the weights, adds the payment feed, or retunes a threshold. The remedy sits on the output channel: for one population, for five years, the composite score and the recommendation do not leave the building, and for the other products the default flips so that an unstated voucher status fails safe rather than fails open. Restoration is conditioned on a named civil-rights organization finding the score valid for that population — a private class settlement installing a civil-rights organization as a gatekeeper over a commercial model's output, with a federal court holding enforcement jurisdiction behind it. That is a governance shape worth studying on its own terms, and it should be read with its bounds visible. Suppression shifts the decision input rather than eliminating it: the underlying credit and eviction records still reach the landlord, and a third-party credit score may still pass through with its source disclosed, so what changes is which number carries the weight. The merits were never adjudicated; the company admitted nothing and maintains its scores comply with law. And as of mid-2026 no compliance report, no enforcement motion and no validating-organization examination had surfaced publicly — the gate exists on court-approved terms and has never been observed operating. The people this all lands on are outside every model here: the eventual harms are a costlier apartment in a worse neighbourhood, a 120-day voucher clock consumed by a denial, and staff time at a legal-aid office diverted case by case, none of them computed from any diagram.
The concepts used in this reading are defined in the Field Guide; the governance responses live in the Practice Library.