Can you sue the algorithm’s maker? Mobley v Workday and the hunt for hidden bias tests

For years the awkward question about AI in hiring has been: if the tool discriminates, who is responsible, the employer that used it or the vendor that built it? In June 2026 a US court moved that question forward.

The case

In Mobley v Workday, the court allowed core discrimination claims against the software vendor to proceed, including claims framed around age and disability, and earlier let a large group of affected applicants join the action. It is the leading test of whether the maker of a hiring algorithm, not just the employer, can be held to account.

The twist

A separate ruling shielded the vendor’s own bias-testing material as privileged. So the very evidence that would show whether the tool discriminates may be hard for claimants to see. That tension, between the push for accountability and the instinct to keep testing confidential, is where a lot of AI ethics now sits.

What it means in practice

Two things follow. First, vendors of hiring AI are increasingly in scope, not just their customers, a point already written into California’s rules through the idea of the employer’s “agent”. Second, testing done well, independently and on the record, is more useful than testing kept in a drawer. The point of an adverse-impact analysis is to find and fix problems early, and to be able to show your work.

We run that testing for both employers and the vendors who supply them. See how one dataset can cover several jurisdictions.

Sources: Mobley v Workday analysis (Duane Morris)

This article is general information, not legal advice.