AI Hiring Tools: In Australia, You Own the Bias, Not the Vendor

Let an AI tool screen your job applicants and something quietly changes hands: the software makes the call, but you keep the liability. If that tool filters out a candidate on a protected attribute — age, gender, race, a disability it inferred from a gap in a CV — the discrimination claim lands on your business, not the vendor who sold you the model. On 20 July 2026, the Victorian Government put employers on notice that it wants to write that expectation into law.

Victorian Labor announced on 20 July 2026 what it called the toughest workplace surveillance laws in the country, to be introduced if it wins the November state election. Two of the measures matter to anyone hiring: employers would have to ensure AI used in recruitment, promotion and pay decisions is not biased or discriminatory, and a human would have to review significant automated decisions that affect a worker's rights. As the law firm Landers & Rogers sets out, the proposal follows a parliamentary inquiry whose final report the government has already backed in principle. It is a proposal, conditional on re-election — not law yet.

Here's the part that doesn't wait for an election. Under the Fair Work Act 2009 and federal and state anti-discrimination law, an employer is already responsible for a discriminatory hiring decision even when an algorithm made it without a human in the room. And the risk isn't the obvious kind. It's indirect discrimination — where a neutral-looking criterion quietly disadvantages a protected group at scale — which is precisely what an untested screening model tends to produce. An Australian employment-law analysis notes that a business relying solely on automated decision-making would struggle to discharge the reverse burden of proof a discrimination claim puts on it. The tool doesn't get called to the tribunal. You do.

This is the same principle we've written about with AI in professional work — you own the mistake, not the AI. Hiring is where it bites hardest, because the harm is done before anyone is employed and the evidence is a dataset, not a conversation you can explain away.

AI has moved fast into recruitment. In a ResumeBuilder survey of business leaders, most companies already using AI in hiring apply it to screening applications — a global figure, but Australian adoption is climbing behind it, and climbing fastest exactly where hiring is hardest. With the country 83,000 tradies short and professional firms chasing the same scarce staff, a tool that reads a hundred applications in a minute is genuine relief. The relief is real. So is the exposure that rides in with it. If you've handed part of your shortlisting to software and can't say how it decides, that's worth a look before your next hire.

Want to know what this means for your business? We'll map it in one call.

You don't need to go back to reading every application by hand — you couldn't keep up, and that's the point. Be deliberate about the one decision you can't delegate: who gets rejected, and why. Keep a human at that call, so the AI ranks and summarises but a person signs off on anything consequential. Ask the vendor a plain question — what was this tool tested on, and has it been checked for adverse impact on protected groups? Avoid screening on proxies that stand in for protected attributes: postcode, a career break, the age implied by a graduation year. And keep a record of why each shortlisted candidate advanced or didn't. That is the human-in-the-loop the Australian Public Service Commission already asks of federal hiring; a private employer that adopts it now is simply ahead of the law. Not sure where AI already sits in your process? Map where it touches your hiring before a claim — or a new bill — maps it for you.

Get this seam right and AI does what it's good at — clearing the volume — while the judgement, and the defensible record behind it, stay where the law already puts them: with you. The businesses that come out of this well won't be the ones that automated hiring fastest. They'll be the ones that knew which decision a machine was allowed to make, and which one a person still had to own.

Key takeaways

Common questions

Related notes

Not sure whether your hiring tools could land you a discrimination claim?

We map where AI already touches your recruitment, mark the decisions that still need a human to own them, and flag the screening criteria most likely to draw a claim — one straight conversation about where the risk actually sits.