Author: Shavi Sharma, Chanderprabhu jain College
Abstract
Workplaces today are becoming a mix of human and machine decision-making AI isn’t just helping out anymore, it’s actually making calls. It screens resumes, rates performance, suggests who gets promoted, and sometimes even flags who gets let go. The problem is, laws written decades ago never accounted for a computer making these calls instead of a person. So now courts and regulators are stuck asking: if an algorithm discriminates, does it break the same civil rights laws that would apply if a human did it? This article looks at how that’s playing out in practice the real court cases, how regulators are responding, and what companies need to do to stay on the right side of the law when they let AI make decisions about people’s jobs.
To the point
HR departments have rushed to bring AI into recruitment, scheduling, performance management, workforce analytics basically every corner of the employee lifecycle much faster than the laws around employment could keep pace. Companies themselves describe the environment they’re operating in as fast-moving, unpredictable, and constantly shifting, and AI isn’t just changing what individual employees do day to day. It’s actually reshaping how management itself works. Layers of hierarchy are getting flattened out. Routine oversight the kind a manager used to do by simply watching people work is increasingly being handed off to software. And managers themselves are spending less time directly supervising people and more time reviewing and interpreting what the algorithm has already decided.
The case for doing all this is easy to make on paper. An AI screening tool can process a volume of applicants that would be completely unrealistic for any human recruiting team thousands of resumes in the time it’d take a person to read a handful. Predictive analytics can flag things like an employee who’s likely to quit soon, or a performance issue that’s about to become a real problem, often before anyone would have noticed it themselves.
But here’s where it gets complicated: the legal side hasn’t caught up to any of this, not even close. If a human recruiter happens to be biased, that might affect a few dozen or a few hundred people over the course of their career. But if an AI system has that same bias built into it, it can affect millions of applicants at the exact same time, in exactly the same way, over and over again. That’s a completely different scale of problem and it’s one that regulators, courts, and lawmakers are really only just beginning to understand and try to catch up with.
The proof
The Black Box Problem
AI hiring tools don’t come with a manual that says “here’s exactly why we rejected this person.” The company that built the tool usually treats the inner workings as a trade secret nobody outside the company (not even the people being rejected) can see what factors the algorithm weighed or how it made its call. So when someone suspects they were unfairly screened out, they can’t just look at the tool and see the bias it’s genuinely hidden inside the system. That’s why lawyers usually have to go through the court process just to force the company to open up the “black box” and show what’s actually happening inside.
Proving Bias Without Proving Intent
In older discrimination cases, you often had to show that someone meant to discriminate like a hiring manager who admitted (or was caught) rejecting people because of their race or age. That’s really hard to do with a machine, because a machine doesn’t have intentions. So the law shifted the focus: instead of asking “did they mean to discriminate,” it asks “did this tool end up hurting one group more than others.” If you can show that pattern with numbers say, the AI rejects women or older applicants at a noticeably higher rate than everyone else that alone can be enough to build a legal case, even if no one ever wanted that outcome.
Use of legal jargon
Disparate Impact : A neutral tool/policy that ends up harming a protected group, even without intent to discriminate.
Agent Liability : Holding a vendor (not just the employer) responsible, since its tool acted on the employer’s behalf in making decisions.
Title VII / ADEA / ADA : Core U.S. anti-discrimination laws covering race/sex/religion, age (40+), and disability, respectively.
Discovery : The legal process forcing a company to hand over internal records, code, or data as evidence.
Class Certification : Court approval to combine many individual claims into one large lawsuit.
Case laws
1. Mobley v. Workday, Inc. (2024)
Facts: Derek Mobley, Black, 40+, with anxiety/depression, applied to 100+ jobs through employers using Workday’s AI screening tool : rejected every time. He sued Workday directly, not the employers.
Significance: First major test of whether a software vendor (not just the employer) can be sued for AI hiring bias. Workday’s tool is used by thousands of companies, so the ruling set a precedent for the whole industry.
Judgment: Court allowed the case to proceed, ruling Workday could be sued as an “agent” of employers : since the AI was making the actual decisions, not just assisting. In 2026, the court extended this to outside applicants too. ~1.1 billion applications were processed through Workday, so the case could become massive.
2. Kistler v. Eightfold AI Inc. (filed Jan 2026)
Facts: Plaintiffs claim Eightfold secretly scraped data on 1B+ workers, scored them 0–5, and silently rejected low scorers before any human review.
Significance: Not about bias : about secrecy and consent. Shifts AI employment law into data privacy/transparency territory, not just discrimination.
Judgment: Still ongoing : court is deciding on class certification; discovery is underway. No final ruling yet.
Conclusion
The human-machine hybrid workplace isn’t a future trend anymore it’s already here, and the law is racing to catch up. Courts are making it clear that handing a decision over to an algorithm doesn’t hand away legal responsibility for it the same civil rights laws that applied to human decision-makers are being applied to AI systems too. For companies, this means AI can’t just be treated as a plug-and-play efficiency tool; it needs auditing, documentation, and human oversight built in from the start. For workers and applicants, it means the systems quietly deciding their job prospects are, for the first time, being held to real legal scrutiny.
FAQs
Q1: Can a company be sued if it didn’t build the AI tool itself?
Yes. Courts have allowed both the employer and the software vendor to be sued, since the vendor’s tool was actually making the decisions.
Q2: Does someone have to prove the AI “intended” to discriminate?
No. Most cases rely on disparate impact proving unequal outcomes for a protected group is enough, intent isn’t required.
Q3: What laws currently apply to AI hiring decisions?
Existing federal laws Title VII, ADEA, and ADA are being applied to AI, along with growing EEOC enforcement and emerging state-level AI rules.
Q4: Is it illegal to collect and score someone’s data without telling them?
It can be : the Eightfold AI case argues undisclosed scraping and scoring is a legal problem on its own, separate from any proven bias.
Q5: What can someone do if they think an AI tool unfairly rejected them?
File a charge with the EEOC (or a state agency) :that’s usually the first legal step before any lawsuit can move forward
