What Happened
Mary Louis, a Black woman in Massachusetts, was denied an apartment in 2021 based on a SafeRent algorithmic score. A class action alleged the scoring system discriminated by race and income: it leaned heavily on credit history while ignoring the reliability of housing vouchers, disproportionately shutting out Black and Hispanic applicants. SafeRent settled in November 2024 without admitting wrongdoing.
Impact
SafeRent paid roughly $2.2 million and agreed for five years to stop showing screening scores or accept/deny recommendations for voucher-holding applicants, with any future scoring model requiring third-party validation. It was one of the first successful legal challenges to algorithmic tenant screening.
Cost: $2.2 million settlement
How to Prevent This
- Test scoring models for disparate impact across race, income source, and voucher status
- Exclude or reweight features (like raw credit score) that proxy for protected characteristics in housing decisions
- Have independent third parties validate screening models before deployment
- Give applicants the specific reasons for adverse decisions and a path to contest them
- Account for guaranteed-payment programs such as housing vouchers in risk models