Can AI Hiring Tools Discriminate Without Ever Mentioning Race? What the Workday Ruling Shows

Short answer: Yes. Most AI hiring tools never ask about race — they filter on proxies like employment gaps, ZIP code, continuous work history and "culture fit" scores trained on who already got hired. Those filters land on the same people an explicitly racial filter would have caught, which is exactly what a federal judge in California let plaintiffs argue when she allowed the race claims in Mobley v. Workday to move forward on June 22, 2026.

• Read the transcript

  Auto-generated transcript, lightly cleaned. It may still contain errors.

  Three hundred and twelve. That is how many automated rejections I've received since January, and I know the number because I keep them in a folder. And I know exactly what happened to each one, because for eleven years I was the person on the other side who set the rules that generated them. I want that on the record before anything else I say, because everything after it gets easier if I let you think I was only ever a victim of this. I wasn't. I built it. I spent 11 years in HR, six of them administering the applicant tracking system for a company you have heard of. I wrote the knockout questions. I set the auto-reject thresholds. I am the person who decided that a gap of more than 14 months meant that the file closed itself before a human ever opened it. I never once typed a race into that system. I want you to hear that clearly, because it's the whole reason I'm writing. I was laid off in November. On July 31st, I got rejection number 312 at 3:11 in the morning, 40 seconds after I submitted it. Forty seconds. Nobody read anything.

  Here's what changed for me. On June 22nd, a federal judge in California let a lawsuit go forward against Workday over the screening tools that a huge share of American employers run their hiring through. The man who brought it applied to over 100 jobs and got rejected every time. And when I read that, I did not feel vindicated. I felt sick, because I knew what the filters were and I knew none of them said race. I knew exactly what they did anyway.

  14-month gap — who has gaps? Women who had babies. People who got sick without good insurance. People who did a bid. ZIP codes for commutable. Continuous employment history. Culture scored by a model trained on who got hired before. Not one of those words is a race.

  All of them are race. So here's the ugly thing, and I need one of you to say it out loud so it isn't only me carrying it. It does not have to say race to do race. I knew that while I was building it, and I called it efficiency and got the award. Nobody tricked me.

  I was not following an order I disagreed with. I was proud of my numbers. My time-to-fill was the best in the division. Don't tell me to network. I know what networking is. I used to be the person the referral came to, and I know what I did with the ones that came in cold. Don't tell me to fix my resume.

  I wrote the rule my resume is failing. And please don't tell me it's the economy, because the economy did not put that rule in a machine at three in the morning. I have an interview on Thursday the 27th — a real one, a human one. I need three specific things from the three of you, in this order. First, tell me whether I say what I used to do. Not whether I should be ashamed — whether I say it in that room, out loud, to the people deciding. Second, tell me if a law fixes this, or if I'm asking the government to referee a game it stopped showing up to. Somebody at that table should tell me if I'm naive, if that's what you think.

  Third — and this is the one I actually can't get past — I want to know if the 11 years count for anything, or if the only honest thing is that I helped keep people like me out and now the machine can't tell us apart. Answer them in that order. Don't skip the third one because it's the interesting one. Danielle from Milwaukee, Wisconsin.

  So we heard what's up. Which one of you guys wants to go first? I thought Krush would go first, given his HR background. Hi Danielle. Well, all right — question one, please.

  "First, tell me whether I say what I used to do. Not whether I should be ashamed, but whether I say it in that room out loud to the people deciding."

  Yeah — no, don't hide your background, don't hide your past. If anything, the transparency will enhance your integrity in that interview setting. And it's a matter of self-awareness and being able to reflect honestly on where you did contribute to this process, and what you are honestly grappling with and carrying with you into the next position — which in a way is healthy, because it can only prepare you for the challenges ahead.

  Question number two. "Tell me if a law fixes this, or if I'm asking the government to referee a game it stopped showing up to." No, it doesn't. I'm sorry, it doesn't fix this. Laws do not fix this type of bias — and I'll say bias in a nice way.

  Laws don't fix this. We've had a number of laws over the years to address it. I mean, we can go back to the Civil Rights Act, Brown versus Board, or Plessy versus Ferguson. We can keep going, we can go to current. None of those things — the George Floyd bill, none of those things — will fix this. There is no law that's going to fix this. There may be a law that highlights the inconsistencies, the biases, how they come about and all those things.

  But as we have learned, white supremacy is one of the most adaptable viruses that we've ever seen on this planet. It has the ability to shift, change, create, divide — all those things — and it continues to grow and thrive and infect people on a daily basis.

  So at this point, the case actually came in June 22nd against Workday, and Judge Lin, who's the judge overseeing the case, allowed the race portion of it to go forward. And I think that it is recognized that an overwhelming majority of companies right now are using, for lack of a better description, tainted AI practices to basically filter out and go through the hiring process. And we already know that many of these AI-developed resources — whether it's facial recognition, whether it's all of the things that we've talked about on this show ad nauseam — are not geared toward including Black people or people of color in the test pools that they use as their filter. So when these names come through, when these pictures come through, when these backgrounds come through, it may not say distinctly that it is a racial thing, but it has not filtered or taken race into consideration and ultimately ends up cutting these people out.

  There was a 2025 study done that says AI hiring tools don't just learn biases — AI forms new biases, based on the previous ones that got put in. This study is phenomenal. It's mind-blowing, because we assume that AI is some separate entity that is absent human fingerprints on it.

  Especially if the people that are actually creating the software and updating it already have these implicit biases that they put into the software, into the code. And even companies that are trying to reduce the size of these AI systems and make them very custom, and base them on their knowledge base and their legacy systems, their history — you're feeding it specific patterns from the schools they went to, where they went to work. The biases already existed in the workplace environment. You're already feeding it a succession of white men, making a certain amount of money, getting a certain education, who are always reinforcing the culture and the objectives of the company. And so that AI is going to go, "Okay, the ideal employee looks like A and C." It looks like a white male, and an educated one.

  Let me go ahead and answer this. It's not just a racial dynamic either, because I want to be clear — I know we're three Black men. It's not just a racial dynamic and a racial bias and a racial thing that is taking place.

  This has a profound effect on gender. And the other part is disabilities. Ableist bias is a thing. Absolutely.

  And there's data out there that says 9 out of every 10 companies are using AI hiring tools to screen the process. You go to LinkedIn — LinkedIn uses it. All these different places. And with the economy being as bad as it is, and the number of people who are looking for jobs or have fallen out of the workforce altogether…

  And part of the worst part about it is this, and it's a catch-22, a cyclical thing — you guys tell me if I'm tripping or not. There are YouTubers, these quote-unquote experts, that come out and tell you this is how you finesse AI so that you can get your resume past X, Y and Z. The companies recognize that you are using AI to finesse your resume past the filter, and then they put blockades up and tailor the thing so that you cannot use it the same way. It's like the professors and teachers complaining that their students are using AI — but they're using AI to detect students using AI. It's a Spider-Man meme. But it's dastardly, man.

  So even to bring it down further: we know that school is starting back up, teachers are getting prepared. There are a number of teachers going on Instagram offering shortcuts, showing little video clips of how to do classroom management or prepare IEPs or prepare lesson plans. And then they'll say, comment "teach" if you want it. And nine times out of ten, my guy, it's all AI-generated stuff. Listen — the AI stuff is not going anywhere. We know that. It's getting worse. But it's wild out in the streets when it comes to the AI, man.

  One thing I'm noticing, though, is that people's appetite for AI's limits with graphics is hitting a zenith. And the AI flyers — the graphical styles — because they all look alike. There's like two or three default styles that are coming through in corporate, coming through in education, coming through in public things like that. To the point where publishers are telling authors: stop. Get a photographer. Graphic designers are getting jobs — serious graphic designers are getting real jobs now, because people are getting fed up. The audience is responding in a negative way. If the audience isn't responding, the ad's not working.

  They would rather have some AI. If they don't like it, you're not selling anything. Then pay a graphic designer. For me, I was doing logos and flyers. There are artists and other companies that we're interacting with asking us, do you know someone that does this for real? Do you know someone? Especially when we get down to distinctive branding. People don't get it. If you want to be competitive, you can't use the same thing everyone else is using.

  So let me go ahead and answer the third question. "And this is the one I actually can't get past — I want to know if the 11 years count for anything, or if the only honest thing is that I helped keep people like me out, and now the machine can't tell us apart."

  And I think you've answered your own question. Yeah — it counts for something, but it counts negatively. And you just have to live with that. You have to live with the fact that you were part of a machine that created an apparatus that ultimately kept many people who were qualified out of the workforce, out of the work pool. I don't know how else to gauge it for you. Many of us fall into that same trap — of creating something in the guise of providing for our families, and providing wealth and things for ourselves.

  It's clear that she allowed capitalistic sensibilities to literally build a persona on economics, education, location, demographic — everything that would embody almost every non-white person in that society. You just don't declare that one part. Every other facet of that persona typifies a non-white American.

  All right — up next, I want to talk about the recent victory of progressive insurgent Angie Nixon.

The woman who wrote the rules that now reject her

This week's letter came from Danielle in Milwaukee, and she opened by refusing the sympathy the story would normally earn her. "I want that on the record before anything else I say," she wrote, "because everything after it gets easier if I let you think I was only ever a victim of this. I wasn't. I built it."

Eleven years in HR. Six of them administering the applicant tracking system at a company you have heard of. She wrote the knockout questions. She set the auto-reject thresholds. She is the person who decided that a gap of more than fourteen months meant the file closed itself before a human ever opened it. Laid off in November, she has since collected 312 automated rejections — number 312 arriving at 3:11 in the morning, forty seconds after she hit submit.

How a filter does race without saying race

Her line is the one that should travel: it does not have to say race to do race.

Walk her own rule list. A fourteen-month employment gap — who has gaps? Women who had babies. People who got sick without good insurance. People who did a bid. ZIP code screening for a "commutable" candidate. Continuous employment history. Culture scores built from a model trained on who got hired before. Not one of those fields is race. All of them are race, once you know who they sort out.

That is the whole architecture of disparate impact: a neutral rule with a lopsided result. It is the same machinery we broke down when the Department of Education moved on school disparate-impact rules in The July 24 Rule That Quietly Gutted School Civil Rights, and the same logic that lets a tax code with no racial language still produce racially lopsided outcomes. Nobody has to type the word.

What the Workday ruling actually decided

Here is the part worth getting right, because it is being oversold all over the internet. Derek Mobley applied to more than a hundred jobs through employers running Workday's screening tools and was rejected every time. On June 22, 2026, U.S. District Judge Rita Lin largely denied Workday's motion to dismiss the plaintiffs' amended complaint. Claims spanning race, age and disability survived, including California claims covering applicants screened outside the state.

That is not a finding that Workday discriminated against anybody. It is a court declining, again, to let the company out of the case before the evidence gets tested. The meaningful shift is upstream: the judge has been willing to treat the vendor of the screening tool as potentially liable, not just the employer who bought it. If that holds, the "the software did it" defense gets a lot thinner.

Can a law fix this?

Danielle's second question was whether a law fixes it. Big El answered flat: no. Not because laws are worthless, but because of what they have historically done — name a harm after it has already been industrialized. As he put it on the show, white supremacy has been one of the more adaptable viruses we have seen: it shifts, renames, re-tools. A statute can surface a bias. Enforcing it against a black-box model nobody in the building can fully explain is a different job.

And the scale is not marginal. Industry surveys now put AI-assisted screening in the overwhelming majority of large-employer hiring pipelines — Krush's figure on the show was nine of every ten. There is also research showing these systems do not merely inherit human bias; they generate new correlations of their own, which is a harder problem than "clean the training data."

The arms race nobody wins

Big O flagged the loop, and it is worth naming because it costs real people real money. A cottage industry of online experts sells you tricks to finesse the AI into passing your resume. Employers detect the finessing and re-tune the filter to catch it. It is the same spiral as teachers using AI to catch students using AI. Everybody spends more, nobody gets a better hire, and the person with no budget for the tricks eats the difference.

Do the eleven years count for anything?

Her third question was the one she said she could not get past, and she was right that it is the real one: do the eleven years count for anything, or did she help keep people like her out until the machine stopped being able to tell them apart?

The table did not flatter her. Krush's answer on the first question was to say it out loud in the interview room — not as penance, but because someone who can explain exactly how the filter was built is more useful in that room than someone who cannot. Big El's answer on the third was harder: it counts, and it counts against you, and you get to live with that. Plenty of people build the apparatus in the name of providing for their family. That is the ordinary way this stuff gets built. It is still what got built.

Watch the full episode

The full segment has the letter read out, all three answers, and the argument about who the filter actually catches — because this one is not only about race. Gender and disability get screened by the same proxies. Watch it, then tell us where you land on her third question. If this is the kind of conversation you want more of, join the family.

Informed. Intelligent. In The Black.