New Jersey Didn’t Write a New AI Hiring Law. It Just Told You Your Old One Already Applies.

New Jersey didn’t pass a new AI hiring statute in 2025. Its Attorney General and Division on Civil Rights did something that changes the compliance math more, not less: they issued guidance stating that the New Jersey Law Against Discrimination — a civil rights statute signed in 1945, the first of its kind in the country — already applies to algorithmic hiring tools the exact same way it applies to a human recruiter who discriminates. No new bill. No new agency. No grace period while a legislature debates definitions. If you’re screening candidates into New Jersey-based roles from a delivery center in Bengaluru, Pune, or Hyderabad, the law you were already supposed to be complying with just got a lot more specific about what “complying” means for an AI screen.

The guidance, in plain terms

The core line from the NJAG guidance: “the LAD applies to algorithmic discrimination in the same way it has long applied to other discriminatory conduct.” That’s a deliberately unglamorous sentence, and it’s doing a lot of work. It means an employer doesn’t get to argue that a biased outcome from a screening tool is somehow a different category of legal problem than a biased outcome from a human interviewer. The Law Against Discrimination’s protections now explicitly extend to automated decision-making across employment — and a Rutgers survey the NJAG cites found 63% of New Jersey employers already use one or more tools to recruit applicants or make hiring decisions. This isn’t a guidance document aimed at a hypothetical future. It’s aimed at what’s already running in production at most New Jersey employers today.

You don’t get to blame the vendor

The part of this guidance that should actually change how a GCC or BPO team reads its vendor contracts: an employer can be liable under the LAD without discriminatory intent, and without fully understanding how the screening tool arrived at its outcome. Liability doesn’t transfer away just because a third party built the technology. If your organization deployed the tool, you own the outcome — the same way you’d own a biased outcome from an internal recruiter, even if you never trained that recruiter to be biased on purpose. The NJAG’s guidance names three places this risk gets introduced: design (an algorithm built on skewed assumptions), training data (data that encodes institutional inequities from years of past hiring), and deployment (a tool used in a context or at a scale it wasn’t built for). A screening vendor can get the first two right and an employer can still create the third one on their own, simply by turning volume up past what the tool was validated against.

Why this is a GCC problem, not just a New Jersey problem

We’ve written before about the four different state theories US employers now navigate since the EEOC pulled its federal AI hiring guidance. New Jersey adds a fifth, and it’s the one most likely to catch India-based screening teams off guard, because it doesn’t look like new regulation. There’s no bill number to track, no effective date circled on a compliance calendar. It’s an interpretation of an existing statute, which means it’s already in force, and it’s the kind of thing that shows up in a plaintiff’s filing before it shows up in a vendor’s compliance deck. If your delivery center screens candidates for a pharma, financial services, or insurance client headquartered in New Jersey — and a meaningful share of GCC clients are — this guidance is the current legal standard for that work, not a future one.

What actually mitigates the risk

The NJAG’s recommended measures aren’t exotic: assess and test the tool before it goes live, evaluate it regularly once it’s deployed, and work directly with the tool’s developer to understand how it was validated rather than taking a sales deck’s word for it. None of that is a checkbox exercise you complete once. It’s closer to what we’ve argued your legal team will actually ask when an AI screening rollout finally crosses their desk: not “does the tool work,” but “can you show me why it made this specific decision, and can you show me that pattern holds across your whole candidate pool, not just the ones who didn’t complain.” A vendor who can’t produce that on request hasn’t been tested — they’ve been trusted.

The uncomfortable read for anyone hoping “no new law” means “no new obligation”: guidance interpreting an existing statute is arguably harder to prepare for than a new bill, because there’s no lead time built in. New Jersey employers, and the GCC teams screening on their behalf, are already covered. The only question is whether anyone checked before this guidance made it explicit.

If you’re screening candidates into US-facing roles from India and want to know what an audit-ready screening record actually looks like under scrutiny like this, talk to us.

Sources: NJ’s Anti-Discrimination Law Applies to Automated Decision-Making — SHRM, Guidance For Employers on Algorithmic Discrimination in New Jersey — Porzio, Bromberg & Newman, 80th Anniversary — NJ Office of Attorney General, Division on Civil Rights