A tenant screening algorithm just cost a company $2.275 million. Not because it was broken. Because it worked exactly as designed. Two Black women in Massachusetts applied for housing with federally funded vouchers. They had guaranteed rental income. A scoring tool rejected them anyway. Why? The algorithm leaned heavily on traditional credit history while completely ignoring voucher income. It treated subsidized renters as high risk when they were statistically among the most reliable. The court ruled something that changes everything for AI companies: if a landlord relies on your score to make decisions, you share liability for discrimination. The "we just make the software" defense is gone. Here's what keeps our team up at night about this case. The algorithm wasn't using race as a variable. It didn't need to. Credit history already carries decades of systemic inequality baked right in. The median credit score gap between White and Black consumers is over 100 points. When you feed that data into a model without deeper intervention, you're not predicting risk. You're automating exclusion. The settlement now requires independent civil rights experts to validate the algorithm before it can issue any recommendations for voucher holders. That's not a patch. That's a fundamental redesign. We think this is where regulated industries are headed across the board. Not just housing. Lending, insurance, hiring. Honest question for anyone building or buying AI tools: who in your organization is responsible for proving your algorithms don't discriminate? #AlgorithmicFairness #AIGovernance #DeepAI