AI job applicant screening algorithms sort queues and foreclose futures – the job not offered, the career not begun, the income not earned – and it does so faster and more finally than any human panel ever could, writes Annalinde Nickisch.
SHE APPLIED on a Tuesday. The rejection arrived on a Thursday. Polite. Instant. Final.
What she never learned was that no one read her application. A model ranked her among the “poor fit” candidates, and the file closed before a human ever opened it.
Somewhere in Melbourne, a woman returning after two years of parental leave is marked down for an “employment gap.”
Somewhere in Sydney, a candidate in her late fifties is filtered out before the shortlist — not because the model sees her age, but because it learned from years of hires who were mostly young, and now favours the pattern they left behind.
Somewhere in Perth, a man whose speech was slowed by a stroke is scored “low communication” by a tool that measures fluency — pace, pauses, the gaps between words — and cannot tell a difference in speech from a deficit in ability.
Three rejections. Three human beings. And not one person who will say: I decided that.
The vanishing decision-maker
When a manager turns you down, there is at least someone to answer for it — a name, a reason, a person who can be asked why.
When a machine turns you down, the reason dissolves. The employer says the system chose. The vendor says the employer configured it. The model, being a model, says nothing at all. Responsibility scatters into the space between them, and the candidate is left holding a rejection that no one will own.
That is the seduction of automated hiring. It doesn’t just filter people. It filters accountability.
You cannot outsource the blame
Except the law doesn’t accept that — not here.
Australia has no AI-specific hiring statute, and it doesn’t need one to reach this. Our anti-discrimination laws are technology-neutral: the Sex Discrimination Act, the Racial Discrimination Act, the Disability Discrimination Act, the Age Discrimination Act and their state equivalents apply whether the party rejecting you is a person or a script.
And they turn on outcome, not intent. If the machine’s decisions fall harder on women, on older workers, on people with disability, the employer can be liable — even if no one meant it, even if no one can explain how the model reached its verdict.
The comforting belief that liability shifts to the vendor is a fiction. It stays with the employer who chose to let the machine decide.
Under the Fair Work Act’s general protections, a candidate knocked back for a discriminatory reason may have a claim regardless of whether a human or an algorithm did the knocking back — and an employer relying solely on the tool must contend with a reverse onus, obliged to prove the absence of a reason it may not be able to see inside the model to find.
So the accountability was never really gone. It was only hidden — and it sits exactly where the machine’s owners hoped it wouldn’t.
The gap where the candidate falls
Here is the cruelty of it: the law protects, but the person it protects cannot see the harm.
An employee has rights to know how they are monitored and managed. An applicant has almost none. In Australia, a private employer is under no general obligation to tell you that AI screened your application, or to explain how it ranked you, or even to admit a machine was involved at all. You cannot appeal a decision you were never told was made. You cannot challenge a bias you were never allowed to see.
The people most exposed – those at the very start, chasing the job that sets their income, their progression, their security – are the ones the system keeps most in the dark.
And the bias comes built in
The machine is not neutral, because its training was not neutral.
It learns from the workforces of the past and inherits their preferences: who “looks like” a leader, who “fits,” whose résumé follows the expected arc. Feed a model decades of decisions made by a narrow group, and it will reproduce that narrowness – at scale, at speed – then present the result as objectivity, laundered through mathematics.
We have seen where this leads. Sold as the cure for human prejudice, the algorithm too often becomes its most efficient distributor.
The rest of the world is already moving
Others have stopped pretending this is hypothetical.
The European Union now classifies AI used in hiring as 'high-risk,' with obligations for transparency, human oversight and documentation, and penalties heavy enough to concentrate the mind.
In the United States, the federal employment regulator has told employers plainly that they answer for algorithmic discrimination – that not understanding your own tool is no defence – and several states now mandate bias audits and impact assessments.
Australia has chosen the lighter touch: apply the old laws, wait and see. There is one crack of daylight — from December 2026, organisations must disclose in their privacy policies how they use personal information in automated decisions. It is a first, narrow admission that being judged by a machine is something you deserve to be told about.
But it is a line in a privacy policy few will ever read, not a right to an explanation, still less a right to challenge the verdict. The distance between us and the rest is widening.
The reckoning
An algorithm does not merely sort a queue. It forecloses a future – the job not offered, the career not begun, the income not earned – and it does so faster and more finally than any human panel ever could.
We were promised the machine would make hiring fairer. Instead, it made the unfairness quieter.
The rejection still comes on a Thursday. The difference now is that no one has to sign it.
Annalinde Nickisch is a business-savvy senior HR management professional with extensive experience and knowledge in Government and Industrial Relations, stakeholder management as well as managing HR operational risks throughout exponential organisational growth.
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Australia License
Support independent journalism Subscribe to IA.
