Six months ago, Charles Lew wrote in Forbes that the era of AI hiring had a legal bill coming due — and he pointed straight at Workday. This week, U.S. District Judge Rita Lin in San Francisco refused to throw the case out. Workday must now defend a nationwide class action alleging its AI screening tools weeded out applicants who were Black, female, over 40, or disabled.
The defense was the exact one Charles Lew flagged in December: “we’re just the software, not the employer.” The court rejected it. Because Workday runs the same algorithmic screening its customers do, Judge Lin held it can be liable as an agent — a de facto employer — under California’s FEHA and the federal ADA. The “we only build the tool” shield doesn’t hold when the tool makes the decision a human used to make.
The sharpest claim left standing is the ADA proxy-indicator theory: the software allegedly flags things like gaps in an employment history and quietly filters those people out — a stand-in for disability that no human ever reviews. That is the “ghosting” Charles Lew described in Forbes: weighed, measured, and discarded by a machine you never see and can’t appeal to. With 80%+ of U.S. employers running these tools, this is the case that decides who answers when the algorithm says no.
Charles Lew’s prediction is now the doctrine on the docket: accountability follows the decision, not the org chart. HITL Score: 22/100 — human oversight at deployment 5/25 (the algorithm screens applicants before a human ever sees them; it is the first and often only decision-maker), ongoing monitoring 4/25 (Workday touts a “Responsible AI program” but fought to keep its internal bias-testing data out of discovery), incident response 6/25 (flat denial — the claims are “false,” the tools “do not make hiring decisions” — rather than disclosure or correction), accountability 7/25 (the bright spot, and the part Charles Lew predicted: the court pinned liability on the party running the decision, so the “software did it” excuse is dying).
