SafeRent scored renters on their credit and never counted the voucher paying their rent

The number that denied Mary Louis

Mary Louis, a 54-year-old Black woman in Massachusetts, found an apartment she could afford. A federal housing voucher covered about 69 percent of the rent at Granada Highlands, and she was ready to sign. The management company, Metropolitan Management Group, ran her through a tenant-screening product and turned her down. The reason was a single figure, her SafeRent Score, which had come back below what the company allowed. When she asked to explain herself, Metropolitan told her it "do[es] not accept appeals and cannot override the outcome of the Tenant Screening."¹

She was not alone in the file. Monica Douglas, a 65-year-old Black woman whose voucher covered about 57 percent of her rent, was denied at a different complex over the same kind of score. Douglas had help from a housing advocate, appealed, and was eventually accepted. Louis was not, and in May 2022 the two of them, with the Community Action Agency of Somerville, sued the company behind the score.¹

The company was SafeRent Solutions, formerly CoreLogic Rental Property Solutions. It sells landlords a risk score, and it had never told anyone how the score was built.² What the case would spend two years arguing was not whether Louis had done anything wrong. It was whether a number could deny an apartment to the applicants least likely to fall behind on rent, and whether the company that sold the number could be held to the Fair Housing Act for it.

What the SafeRent Score counted

A tenant-screening score is a product a landlord buys instead of reading an application. The screening company takes the applicant's name, pulls the records tied to it, and returns one figure meant to predict whether the tenant will pay and stay. SafeRent sold its number as a "lease performance risk" score. According to the United States, which filed a statement of interest in the case, the score "uses an algorithm that considers rental applicants' credit history, other credit-related information, including non-tenancy debts, and eviction history in calculating a score."² Non-tenancy debts are the ordinary financial record that has nothing to do with paying rent: a medical bill in collections, a car loan, a credit-card balance.

Set against that list is the thing the score left out. Louis and Douglas were not carrying their rent alone. A housing voucher covered most of it, and a voucher is close to a guarantee, because the housing authority pays the landlord its share directly each month whether or not the tenant's own finances wobble. The government's summary of the case put the omission plainly: the score weighed factors like credit history and non-tenancy debts "while failing to consider ... that the use of housing vouchers funded by HUD makes such tenants more likely to pay their rents."³

The applicant never saw the line the number had to clear, and the landlord did not weigh it. Metropolitan did not set Louis's score against her voucher or her rental history. It read the figure, told her it was "lower than is permissible," and closed the file.¹

Why a credit score falls hardest on voucher holders

Two facts about those inputs explain the pattern the plaintiffs alleged. The first is about race. Credit scores are not spread evenly across the population. Median credit scores are lower for Black and Hispanic Americans than for white Americans, a gap that reflects decades of unequal access to credit, and a score built mostly on credit history inherits that gap. The plaintiffs' theory was that the SafeRent Score, leaning on credit and non-tenancy debts, returned lower numbers for Black and Hispanic applicants as a group and denied them at higher rates.⁴

The second fact is about the voucher. The applicants most exposed to this were voucher holders, and a voucher is the strongest assurance of payment a landlord can get. The score measured the debts a voucher exists to cover, and never counted the voucher. The number the plaintiffs described treated its safest applicants as its riskiest.

That is a claim of disparate impact: a rule that applies the same measure to everyone but falls hardest on a protected group. Under the Fair Housing Act, a practice can be unlawful for its effect alone, with no showing that anyone meant to discriminate, when it disproportionately harms a protected class and is not justified by a substantial, non-discriminatory reason.² The newer question this case raised was who could be sued for it. SafeRent is not a landlord and rents no one an apartment, and it argued the Fair Housing Act did not reach a company that only sells a score. The United States disagreed, telling the court that the statute "focuses on prohibited acts," not "specific actors," so a screening company that helps decide who gets housing can be directly liable.²

On the substance, SafeRent's position was that no one could see inside the score, its critics included. The company does not disclose its algorithm to landlords, applicants, or the public, so what it actually weighed, and how heavily, is not in the record.² That cuts both ways. The plaintiffs could point to no line of code that down-weighted voucher holders; their case rested on the inputs the company acknowledged and the outcomes those inputs produced. And no court ever tested the statistics. The claim cleared a motion to dismiss, the stage where a judge assumes the alleged facts are true and asks only whether they would break the law if proven. In July 2023 Judge Angel Kelley ruled they would, and let the disparate-impact claim proceed.¹ It never reached a trial.

How SafeRent settled without conceding the score

Judge Kelley's order let almost the whole case survive. She denied Metropolitan's motion to dismiss outright, and threw out only two state-law counts against SafeRent while keeping the federal fair-housing claim and the state race and source-of-income claims alive.¹⁵ With the claims cleared for trial, the company chose to settle. Its account of why never changed.

The agreement, given final approval on November 20, 2024, put the total at $2,275,000, of which up to $1,175,000 goes as cash to the applicants in the class and the rest to fees and administration.⁷ The part that reaches past the plaintiffs is the injunction. For at least five years, SafeRent may not attach an approve-or-decline recommendation from its score to a screening report for an applicant using a housing voucher, unless it first has the underlying model validated for fairness by a third party the plaintiffs accept. Where it does not, it gives the landlord the background information and leaves the decision to a person.⁷

That writes into a contract the one step that saved Monica Douglas and was denied to Mary Louis: a human who can look past the number. Douglas appealed hers and got the apartment. Louis had no appeal and lost it.¹ "Everything is based on numbers," Louis said after the settlement. "You don't get the individual empathy from them."⁶

Where a score still stands in for the landlord

The habit the case interrupted is now the norm. In a November 2022 market study, the Consumer Financial Protection Bureau found that "most landlords appear to screen prospective tenants," and that the largest, corporate landlords lean hardest on automated systems that return a proprietary score in place of case-by-case judgment.⁸ The same report traced SafeRent's own lineage, noting that CoreLogic sold its tenant-screening business, now SafeRent Solutions, in 2020 and 2021.⁸

The weakness the lawsuit named belongs to the products, not to one vendor. The regulator found that "prior rental payment history is overwhelmingly not reflected in reports or algorithmic risk scores assigned to tenants," so the record most directly about whether a person pays rent is often the one the score skips.⁸ And because the scoring is proprietary, it "can obfuscate the underlying reasons for adverse decisions": an applicant a number turns down is rarely told which record sank them, or given a way to fix it.⁸ That was Louis's position, reproduced across a market where, on the CFPB's count, about 68 percent of renters already pay an application fee to be screened.⁸

The plaintiffs' lawyers cast the settlement as a warning to that market. "Management companies and landlords need to know that they're now on notice," said Todd Kaplan of Greater Boston Legal Services, "that these systems that they are assuming are reliable and good are going to be challenged."⁶