- A property manager's screening algorithm and pricing algorithm are two different lawsuits waiting to happen — under two unrelated federal statutes. Fair Housing Act on one side, Sherman Act on the other. Most firms treat them as one problem. Courts don't. 🧵
- Front one: screening. SafeRent's Registry ScorePLUS scored housing-voucher holders low by leaning on credit history — ignoring the guaranteed income a voucher provides. Settlement: $2.275M, Nov 2024. The court killed the "we're just a neutral vendor" defense.
- Why credit-weighted screening is a disparate-impact trap: median FICO runs 727 White, 667 Hispanic, 627 Black. Over half of White households clear 700; only 21% of Black households do. Weight credit heavily and you encode that gap into approval rates.
- The test isn't intent. It's the four-fifths rule: if your approval rate for any protected group falls below 80% of your top group, that's a presumptive Fair Housing Act violation. The algorithm doesn't have to mean it. The numbers convict it.
- Front two: pricing. RealPage's AIRM pulled non-public rents and occupancy from competing landlords and nudged prices "in unison." DOJ called it a hub-and-spoke cartel. Landlord class-action settlements now top $140M — Greystar alone $57M.
- The kicker: AIRM auto-accepted recommendations within a 3% daily / 8% weekly band, and most landlords never touched the defaults. The algorithm was effectively setting rents with no human in the loop. ProPublica pegged the cost at ~$70/month per renter.
- The winning defense already exists. Yardi beat its California case because a court accepted that Revenue IQ "does not and by design cannot use any client's confidential pricing information for any other client." Architecture, not affidavits, won.
- And the map is fragmenting. CA AB 325 (Jan 2026) bans "common" pricing tools used by 2+ firms. NY S.7882 bans coordinating-function tools even on public data — so Yardi's "we isolated it" defense won't save you there. CO SB 205 mandates impact assessments.
- Add agentic leasing. Autonomous agents that pre-screen, steer, and negotiate commit FHA steering and price-signaling acts at machine speed, no human in the loop — the "excessive agency" problem. One platform runs 1 in 12 US units. That architecture doesn't exist yet.
- The two lawsuits share one root cause: the screening team and the pricing team have never met. FHA risk lives in one, antitrust in the other, and no horizontal "AI governance" tool maps both. So — which one lands on your firm first? #HousingAI
- We audit screening for disparate impact, run LDA search on your models, and engineer provable data isolation for pricing. No horizontal AI-governance tool covers both fronts — we built the one that does: https://veriprajna.com/solutions/housing-ai-compliance