Your pricing algorithm never sees race or income. It sees ZIP code, device type, and browsing time — and that's where the legal risk hides. A shopper on an older Android in a lower-income ZIP can get a different price than an iPhone user in a wealthy suburb. It never intended to discriminate. But census data shows those clusters track with race and income closely enough to fail a disparate-impact test. The FTC's $60M Instacart settlement in 2025 made this concrete. Its tool generated up to five prices for the same item in one store, with swings as high as 23%. It didn't hinge on intent but on outcome: certain shoppers systematically paid more. That's only one of two enforcement tracks. The other is collusion: Amazon's Project Nessie raised prices on 8 million items by predicting when rivals would follow, and now faces a 2026 antitrust trial. Run a third-party vendor that also serves your competitors, and Gibson v. Cendyn (2025) decides your exposure: sharing an algorithm isn't illegal by itself, but the moment that vendor markets raising prices industry-wide or pools rivals' non-public data, you're in hub-and-spoke territory. Our take: when an FTC civil investigative demand lands, "we didn't mean to" is no defense — only proof of what your algorithm did, to whom, and why. If that demand landed tomorrow, could you reconstruct why one customer saw the price they saw — or would you be starting months of forensic data extraction? #AIGovernance #AlgorithmicPricing