AI in Hiring Is Already in Court — And HR Hasn't Caught Up

The dominant conversation about AI in HR is still about efficiency — faster screening, leaner operations, smarter performance reviews. The more important conversation is about liability, and it's already happening in courtrooms and employment tribunals, not boardrooms.

A federal court recently dismissed a racial bias claim against Sirius XM Holdings Inc., in which the plaintiff alleged that AI-assisted hiring tools screened him out of approximately 150 open roles — in part over questions about his use of multiple addresses — without any human review of the pattern. HR Dive reported the dismissal, but the win for Sirius XM should not be read as a green light. Courts dismissed the claim on procedural and evidentiary grounds, not because AI-driven screening is categorically lawful. The gap between "we won this case" and "our process is defensible" is exactly where the next lawsuit will find its footing.

What does the Sirius XM AI bias case actually mean for employers?

The answer that most HR leaders want — "we're fine if we win" — is the wrong one. The Sirius XM case illustrates that AI-assisted applicant tracking systems (ATS), tools that automatically rank, filter, or flag candidates based on machine-learned criteria, can generate discriminatory patterns without any individual human intending to discriminate. An ATS, defined plainly, is software that screens job applications before a human ever reads them, using rules or models trained on historical hiring data. If that historical data encodes bias — and most does — the tool propagates it at scale. Winning one lawsuit doesn't audit the system. It just buys time.

Are AI-generated employee grievances a real HR risk or a niche problem?

They are real, and they are accelerating. Personnel Today reports that human resources (HR) professionals are increasingly encountering workplace grievances drafted by generative artificial intelligence (AI) tools — complaints that are more formally structured, legally precise, and voluminous than those typically written by employees unaided. This matters for two reasons. First, AI-drafted grievances are harder to dismiss informally; they arrive pre-loaded with legal framing that forces a procedural response. Second, the volume effect is not trivial — when the friction of filing a grievance drops to near zero, filing rates go up. HR teams that sized their employee relations capacity for the old friction level are already behind.

These two dynamics — AI in hiring generating liability, and AI in employee hands generating process load — are the same story from opposite ends of the employment relationship. Organizations that treated AI adoption as an operational question are now discovering it was always a legal and governance question.

The performance review is where this tension becomes most acute. HR Dive cites Highwire's chief executive officer (CEO) making the point directly: AI-assisted performance evaluations lack the contextual judgment to distinguish a smart risk from a careless mistake. That distinction is exactly what employment tribunals and juries are asked to make when a terminated employee challenges their review. An AI-generated performance narrative, stripped of managerial context, doesn't just risk being wrong — it risks being undefendable in a hearing room.

What's striking is how unevenly organizations have absorbed this. The same employers offering higher-than-planned salaries — more than half of employers, according to HR Dive — to attract candidates with AI expertise are, in many cases, deploying that expertise inside processes that haven't been legally stress-tested. You can buy the talent. You cannot buy the governance infrastructure retroactively.

The Glovo case in Spain, documented by MIT Sloan Management Review, is the cleanest preview of where this leads. Glovo's attempt to work around Spain's 2021 Rider Law — which presumed employment status when a digital platform directs or controls workers through algorithmic management — ultimately backfired, resulting in enforcement action and reputational damage that dwarfed the compliance cost the company was trying to avoid. Algorithmic management tools, whether applied to gig couriers or salaried knowledge workers, are increasingly being treated by regulators and courts as employment decisions, not just software outputs.

The question for chief human resources officers (CHROs) is not "have we been sued yet?" The question is: for every AI-assisted decision in our talent lifecycle — screening, evaluating, managing, exiting — could we reconstruct the human judgment that authorized it? If the answer is no, you have an audit problem waiting to become a litigation problem. The courtrooms are open. The filings are already accumulating.