Erin Kistler spent the better part of two decades building the kind of résumé that used to open doors on contact — product manager, with PayPal and Microsoft and Netflix among the names on it — and then she spent four years watching it vanish into a hole that never writes back. She applied for thousands of jobs in those four years and did not land a single interview. Not one. The reason no one ever explained a rejection is that no person was making the call. A machine scored her on a scale of zero to five, without her knowledge, without her consent, and without ever showing its work. Now she is trying to drag that machine into a courtroom, because that is the one place left where a human being has to answer for what an algorithm did to her.
The vendors who sold algorithmic hiring promised it would strip out bias. What it actually stripped out is the only part of the process where somebody had to tell you why you lost.
Erin Kistler — Twenty Years of Experience, Zero Interviews
Kistler is the lead plaintiff in a proposed class action against Eightfold AI, the Silicon Valley firm whose screening software is used by hundreds of employers — including, she says, the companies where she applied. Filed in January in California, the suit makes a deceptively simple argument: when a vendor scores and ranks you before any human ever reads your résumé, that score is a consumer report about you, and the law that governs consumer reports says you get to see it and challenge it. The Guardian has framed the fight plainly: a rise in lawsuits over AI in hiring is forcing the question of how companies decide who gets in and who gets cut. Eightfold counters that its score is nothing of the sort — just a proprietary ranking. The fight, in other words, is over whether a woman has the right to read the report that has been quietly deciding her life.
Her lawyer, Rachel Dempsey, boiled the stakes down to one sentence: the concept of a black box is very scary. An American can pull a credit report and dispute a wrong line. Kistler cannot pull the number that has been keeping her unemployed, because the company that wrote it says she has no right to look at it.
Eightfold denies the allegations and says it will defend itself vigorously. That is the corporate position. The human position is a woman with twenty years of skill who still does not know what the machine thinks is wrong with her.
Eightfold AI — a Score From Zero to Five You Never Get to See
The scale of this is where it stops being one woman’s lawsuit and starts being a structural fact. Eightfold calls itself the world’s largest self-refreshing source of talent data, and that is not hype: it maintains a database assembled from the résumés, LinkedIn profiles, and social-media trails of more than a billion workers. Every application feeds the machine, and the machine assigns each applicant a score from zero to five predicting how well they will perform in a role. The highest scores surface for interviews. The lowest scores simply stop existing.
The legal opening is narrow and specific. Ogletree Deakins has called the Eightfold case the first real test of whether AI screening tools trigger the Fair Credit Reporting Act — whether a silent score, used to rank and discard applicants before a human ever looks, counts as a consumer report that must be disclosed. If the answer is yes, the entire scoring industry gets handed a transparency requirement it never wanted.
And the baseline is enormous. By the World Economic Forum’s count, ninety percent of employers were already using some form of automation in hiring last year. That is not a trend anymore. That is the default. Which means the quiet zero-to-five number has become the single most important document in millions of working lives — and the only one those workers are never permitted to read.
Ifeoma Ajunwa — You’ve Been Algorithmically Blackballed
Emory law professor Ifeoma Ajunwa, who studies AI and the future of work, points to the detail nobody in the industry wants on the record: there is no law anywhere requiring a company to tell a job applicant they are being evaluated by AI. So companies simply do not. The systems were supposed to strip out human prejudice; Ajunwa’s research keeps finding the opposite — the tools absorb the biases of the managers whose data trained them, then run them at machine scale. Amazon built a screening model that downranked women’s résumés because its historical top performers were men. In her own work she watched an AI ding applicants for southern accents it simply could not parse.
The consequences travel further than a single opening. Because so many employers license screening built on the same underlying models, one negative flag can follow a person from application to application, silently, for years. Ajunwa has a word for what happens to a candidate caught in that loop: algorithmically blackballed. A human rejection lets you try again somewhere new. An algorithmic rejection remembers you — and makes the same decision again for the sake of efficiency.
If you think the machine has merely inherited our flaws, consider the University of Chicago study by Xuechunzi Bai, who fed hiring models fictional candidates sorted into invented groups — Tufa, Aima, Reku, Weki — and watched the machines manufacture stereotypes out of thin air. If one Tufa made a good doctor, every Tufa got routed toward medicine and every Weki toward the mop bucket. The models were not just biased. They were more biased than the human hiring decisions in the same study, and the newer, more capable models were the worst offenders of all. The smarter the machine, the harder it stereotypes.
Meta and IBM — the Machine Has Moved From Hiring to Firing
If the black box starts at the front door, it is already moving downstream. Bloomberg Law has reported on the suit by Meta workers who allege an internal AI system helped select them for layoffs precisely because they had taken medical, parental, or family leave — the machine treating a legally protected absence as a performance problem. IBM faces a parallel suit alleging its AI tools discriminated against older workers. And in the case every employment lawyer is watching, Mobley v. Workday, a federal judge in California ruled in June that a Black applicant over forty — who says he was passed over for more than a hundred jobs — may keep pressing claims that screening software rejected him because of his race, age, and disability. The court’s ruling, handed down by Judge Rita Lin, refused to let the company hide behind the argument that its scoring was merely advice. Workday denies the claims. The case moves forward anyway.
Read the pattern instead of the press releases. Four years ago the machines screened you. Now they fire you. The only question left is whether they will one day decide you were never really in the running at all.
The Fix Is Obvious and Nobody Wants It — Show the Math
There is a remedy, and it is almost embarrassingly simple: make the score visible. New York City already requires employers to run annual bias audits and to warn candidates when automated systems help decide their fate. Illinois and Colorado have followed with laws barring AI tools that produce unlawful discrimination. The loophole is identical in every one of them — the rules bind tools that “substantially assist” a decision, and the vendors keep a human signature on the final line precisely so the software can insist it never decided anything.
Here is the sentence the industry will not say out loud: if the algorithm is only a tool and the human truly makes the call, then the human should be able to explain the call. When a company cannot tell a twenty-year veteran why thousands of applications yielded zero interviews, the human was never making the call. The machine was — and it was never programmed to apologize.
A whole generation of workers is learning what it feels like to be turned down by a wall. The lawsuits, at bottom, are not really about discrimination. They are about a smaller and more radical demand: let me see the report. If a machine is going to decide who eats, the least it can do is show its math.
Sources: The Guardian, Bloomberg Law, Ogletree Deakins, RPJ Law.