Recruitment edition
A lawsuit over 1.1 billion rejected job applications puts your ATS under the microscope, and AI leaders are hitting the brakes themselves.
Workday must hand over bias tests: your ATS vendor could be next
The lawsuit over AI screening keeps growing, and judges say: employer and vendor are jointly liable.
The Mobley vs Workday class action drags on, and the pressure is rising. Judges issued so-called discovery orders in late June and late July 2026. That means Workday has to hand over more and more customer data and bias-test data. Court filings mention a striking number: since September 2020, 1.1 billion job applications have been rejected using Workday's software tools. That makes the possible claims huge. The case is about alleged discrimination by age, race and disability through AI screening. For you, the key point isn't the American courtroom. It's the fact that judges are confirming that employers and vendors can be jointly liable for discriminatory outcomes, even without intent. So you don't need to have meant any harm. What matters is what rolls out behind the scenes. Experts advise organisations not to rely only on the vendor's guarantees, but to audit their own hiring AI. A screening tool learns from past data who counts as a 'good' candidate. If that data mostly shows young people without a gap on their CV being hired, the model repeats that pattern. The tool doesn't ask for age, but it picks up on signals linked to it: graduation year, years of experience, words used in the CV. That's why this is called indirect discrimination: you won't notice it in one rejection, but you will across thousands. The discovery orders centre on whether bias tests exist and what they showed. A bias test compares the pass-through chances of different groups of candidates within the same selection process. A guarantee from your vendor says something about their model, not about your settings, your knock-out questions or your candidate pool. Auditing it yourself is easier said than done, because you need figures your vendor often doesn't provide by default. The practical middle ground: don't stop using AI screening, but record who falls out of the selection and why, and review that yourself. This week, ask your ATS vendor in writing whether bias tests have been carried out, and what they showed.
Background for subscribersThis edition shows you can't just trust your vendor blindly. The Workday case and the ChatGPT ruling show that you share responsibility for what your AI tools do with candidate data, even without bad intent. At the same time, bottlenecks are shifting: a voicebot doesn't solve a problem, it moves it to your calendar. And candidates increasingly form their picture of your vacancy via ChatGPT, out of your sight. In short: check your tools, your sources and your process yourself, instead of relying on guarantees.
AI leaders are hitting the brakes themselves: don't count on ever-faster tools
This weekend, Sam Altman (OpenAI) joined Dario Amodei (Anthropic) and Elon Musk. They're calling for a deliberate slowdown in developing advanced AI, so people keep control. Altman wrote on X that no American competitive pressure justifies recklessness. He warned about two things: people losing control to AI systems, and one company or person using AI to concentrate power. Amodei put forward a three-step plan, including independent overseers who get 'employee-like access' inside the big AI labs. Markets reacted straight away: SoftBank fell 13% on Monday in Tokyo, and Samsung Electronics and SK Hynix dropped more than 4% in Seoul. In practice, 'pacing' means new features arrive later, or in a limited form, because safety checks come first. This has knock-on effects for hiring tools: your ATS or sourcing tool often runs on a model from OpenAI or Anthropic. If that layer slows down, it delays the features your vendor is promising you. That's not a disaster, but it is different from the roadmap sales teams are showing you now. You can read this as honesty, or as a power move: companies calling for rules they already meet put newcomers at a disadvantage. So don't base a hiring process on an AI feature that's only supposed to be 'really good' next year. Base it on what the tool can demonstrably do today.
Background for subscribersThis is a taster.
Subscribers read the whole Recruitment edition: every story, each with its background in plain language.
Subscribe →Not ready for that? Read the free letter first →
AI and tech, made simple.
This is the Recruitment edition. Subscribers get the background to every story, in plain language.
Subscribe
inklaretaal · Amsterdam · © 2026