Is Consent Becoming Fictional?

There are more than 130,000 people in Perth and Fremantle who recently took part in an Australian-first experiment in artificial intelligence, most of them without knowing it. They walked past a camera.

The Western Australia Police have been trialling live facial recognition in public places, comparing faces against a watchlist of about 4,000 people. There were 33 alerts and 19 arrests in the first week, according to the ABC last week. Police say non-matches are deleted almost immediately, and that early results are promising. They may well be. Finding someone wanted for a serious offence is useful. Finding a missing person could be invaluable. There is little point pretending otherwise merely because a computer is involved. The more interesting question is why finding those people requires examining the faces of everybody else.

Facial recognition has already reached Australian shops and stadiums. Bunnings used it to check customers against a database of risk. Some venues now let your face do the work a ticket used to. Each use is different, and each comes with an good reason: crime, staff safety, queues, convenience. Nobody has yet proposed facial recognition because it would be rather nice to know where everyone is. That would sound intrusive, so instead we acquire the machinery one perfectly sensible application at a time.

WA Police displays signs where the technology is operating, which it calls making the surveillance overt. It does. What it doesn't offer is much of a choice: you can be told your face is being scanned, but you can't decline. There is a difference between actual consent and implied consent: one is what you give when somebody asks and you say yes, the other is what the law assumes because you kept walking after being told. One requires your agreement, the other only requires that you didn't argue. Consent has undergone quite an administrative improvement. Retailers have a version of this, and venues another: reject the conditions, and you simply don't enter the shop, or the football, or whichever service wants one more piece of your information. A peculiar sort of freedom.

Facial recognition isn't quite the same thing as CCTV, which only ever recorded what had already happened. This asks a different question, live. Who is that? Once a camera can answer that automatically, you are no longer merely visible, you are searchable. For most people, most of the time, nothing will happen, which is meant to be reassuring.

The standard answer to privacy concerns arrived long before artificial intelligence: if you've got nothing to hide, you've got nothing to fear. It's an admirably economical arrangement, since it places the entire obligation on the person being watched. Privacy has never required criminal intent. We close curtains and shut bathroom doors without that being evidence of a crime. Walking down a street shouldn't require a declaration of innocence either. It's worth remembering that Australia has no general right to privacy written into law, the way some other democracies do. What we have instead is a set of specific protections covering specific things, and a great deal of open ground in between. Facial recognition in public was built for exactly that ground.

People have perfectly ordinary reasons for going somewhere without it becoming part of a record: a doctor, a political meeting, a protest. Or they might simply be doing nothing interesting at all, which deserves protection too. The real change is the presumption underneath it. Traditional policing starts with a person of interest and looks for evidence. Mass facial recognition starts with everybody and asks who might be. The camera looks the same either way.

Much of the debate concerns accuracy, and reasonably so: research shows differing performance across demographic groups, with particular concern in WA about First Nations faces and communities already over-policed. But suppose the technology gets better. Suppose it becomes extraordinarily accurate. A machine that wrongly identifies thousands of innocent people will eventually become politically embarrassing. A machine that identifies almost everybody correctly is considerably more useful. At that point the question is no longer whether facial recognition works. It's whether we want it to.

WA Police's watchlist can include those wanted for arrest, those accused of serious offences, missing people, and people considered a risk to themselves or others. This is where it gets interesting: the camera is only one part of the system. The list behind it matters more, who qualifies, and who decides? How long does a person remain on it? If the information is wrong, how would anyone find out?

These aren't objections to finding dangerous people. They're questions about machinery that tends to outlast the reasons it was developed. Trials become permanent once the vans are purchased and the officers trained. Watchlists grow the same way, stretching from warrants and missing persons to people merely considered "at risk." Nobody voted on that. It's never one dramatic decision, just a hundred small ones, each defensible on its own.

Perhaps the strangest thing about our developing surveillance system is that nobody appears to be building one. The supermarket is reducing theft, the stadium improving entry, the police finding people, each with its own camera, its own database, its own reasonable purpose. The citizen gets the combined result: a dozen systems, separate for now, each capable of a good deal more. This is usually called efficiency.

Other countries have already had this argument. Britain has embraced live facial recognition, arrests cited as proof of success, civil liberties groups fighting it through the courts. The EU took the opposite approach, its AI Act prohibiting real-time biometric identification by police in public in principle, with only narrow, independently authorised exceptions. Australia has neither prohibition nor a case yet before a court. What it has is a trial, running while the debate catches up.

There are circumstances where facial recognition may be justified: an abducted child, an immediate danger. That doesn't settle the argument. It begins it. A democratic society can decide some surveillance is necessary, while deciding the burden of justification belongs with the watcher, not the watched. Who may be scanned, and for what? Who audits the results when the system is wrong? What prevents today's exceptional use becoming tomorrow's ordinary one?

And perhaps one question before all the others: when did being in public become consent to being identified? We have been told for years that if we have nothing to hide, we have nothing to fear. Perhaps. But a free person shouldn't need anything to hide before being entitled to be left alone.