In a first for Australian law enforcement, police in Western Australia have deployed live facial recognition technology in marked vans at locations around Perth.
The system scans the faces of passersby and compares them in real time to a watchlist of around 4,000 people with outstanding warrants. The list also includes registered sex offenders and missing persons. When a potential match occurs, nearby officers are alerted.
Police say the trial of the technology is “a way that we can increase the freedoms and the privacy of our community.” But facial recognition technology has documented drawbacks and risks, and deploying it in this way will present new ones — and Australia’s legal and governance systems are ill-prepared to manage them.
The case for the WA police program
WA police commissioner Col Blanch states the trial “is not about mass surveillance.”
The commissioner says the cameras will lead to better community protection from repeat sex offenders who might be violating restrictions. Other goals are to apprehend serious offenders and find missing persons more quickly.
There are some safeguards in place. No records are kept of any faces not on the watchlist, and the watchlist is limited to serious offenders.
The technology itself is not new. Police have used facial recognition on pre-recorded footage for more than ten years. What is new, however, is using it in real time on people walking down the street.
The case against
Facial recognition technology poses several well-known concerns.
First among these is accuracy. Even state-of-the-art systems make mistakes.
The most popular facial recognition algorithms are estimated to achieve accuracy rates of around 90% under controlled conditions — but real-world deployment is messier.
Even at 90% accuracy, one in ten alerts is wrong. This may lead to a large number of people being stopped, questioned, and potentially detained for things they did not do.
Another concern is algorithmic bias. AI systems trained on data that does not properly represent the full range of human faces may be less accurate for non-white people. In the Australian context, this risks compounding the incarceration rate of Indigenous Australians, already the highest in the world.
Yet another concern is automation bias — the tendency to trust software more than human judgment. In the United States, this has led to several wrongful arrests.
There is also a risk of function creep. While the system has been deployed now to search for 4,000 serious offenders and missing people, once it is in place it may be used for other purposes, such as targeting visa overstayers or protesters.
What does the law say?
In Australia, the primary legislative instrument governing the collection of biometric data is the Privacy Act. This treats facial images as sensitive information, with the highest level of protection.
Ordinarily, people must give consent before such data about them can be collected. But police forces have a different legal position from retailers or employers.
Under Australian Privacy Principle 3, a law enforcement agency may collect personal information from a source other than the individual concerned, if collecting it directly could jeopardize an investigation.
If an enforcement agency believes the collection of information is necessary for its lawful activities, the consent requirement does not apply. In practical terms, that means the public has fewer formal privacy protections against police use of facial recognition than against a retailer doing the same thing.
What are the ethical issues?
Societies always face a tension between security and freedom. Both are good things, but drawing a line when they come into conflict is not always easy.
Even the existence of this technology may change people’s behavior, especially given its imperfect accuracy. If people know they are being scanned, they might avoid public places, or demonstrations, or even religious gatherings, because they are worried about false positives — or because they don’t trust the state’s intentions.
There is also the issue of how the lack of consent involved affects the unwritten agreement between people and government. Even in otherwise democratic societies, some will perceive the use of facial recognition in public spaces as an erosion of civil liberties.
As the first such trial in Australia, the WA program is likely to set a precedent. Whatever standards are established, or not established, will likely flow on to other Australian jurisdictions who are watching with an eye to the future.
The governance question
What would responsible governance of this technology look like?
Facial recognition technology and other biometric scanning are a high priority for the Office of the Australian Information Commissioner, which administers the Privacy Act.
Australia has no dedicated legislation around surveillance using AI or biometric data. Nor does the federal government’s National AI Plan, published in December 2025, recommend introducing any.
Responsible rules for governing WA’s program should require, at minimum, independent audits of field accuracy and algorithmic bias. Clearly defined limits on watchlist scope would also be essential.
There would also need to be someone in charge of enforcing the rules — perhaps a statutory oversight body with genuine enforcement powers. At present, the federal government has established an AI Safety Institute, but it only has advisory powers.
The WA police are not the villains of this story. The problems they are trying to solve, and the people on the watchlist, are not abstractions.
But the law has not kept pace with the technology. The problem is the lack of a robust and nationally accepted legal framework, which means the facial recognition trial will be a poorly controlled experiment in civil liberties.
David Tuffley, Adjunct Senior Lecturer, Applied Ethics and CyberSecurity, Griffith University
This article is republished from The Conversation under a Creative Commons license. Read the original article.
Follow us on X, Facebook, or Pinterest
