In a single year, NSW's AI seatbelt cameras issued more than 126,728 fines, a 2,357% jump. But here is the part that gets drivers fired up: 88% were for a belt worn "incorrectly", not for no belt at all. And "incorrectly" is the hardest thing a camera can judge from a single frozen frame.
In a single year, New South Wales issued more than 126,728 seatbelt fines by AI camera, up around 2,357% on the previous year. Each one carries about $400 and three demerit points. The Guardian reported that seatbelt-fine revenue in NSW jumped roughly 1,400% as the cameras scanned about 140 million cars.
To be fair, some of it is working as intended: camera seatbelt fines fell 12.6% the following year, from 126,728 to 110,738, so more people are buckling up correctly. That is a genuinely good outcome. But it is not the whole story.
Only 12% of the fines were for no seatbelt at all. The other 88% were for a belt worn "incorrectly". That distinction matters enormously, because "incorrectly" is a judgement call, and it is exactly the kind of call a camera makes badly from one low-contrast image.
Picture what these look like. You turn for a second to settle a child in the back and your sash slips off your shoulder. A dark belt vanishes against a dark top. A belt sits behind an arm as someone reaches for a bag. None of these is a person choosing to ride unrestrained, yet each can read as "no seatbelt" in a single frame. Improper wearing can be a real safety issue, so not every one of these is wrong. But it is the most error-prone category there is, and it is where legitimate disputes pile up.
That is the uncomfortable heart of it. There is no published figure for how many of the 126,728 are mistakes, and there cannot be, because most people just pay. An error only ever surfaces if the driver bothers to fight it, so the true number stays hidden by design.
What we do have are proxies, and they are not reassuring. When drivers in Western Australia actually challenged their AI seatbelt fines, around 60% were overturned. Testing of the seatbelt-detection AI put its false-positive rate at 15.28%, more than one in seven. That figure is measured at the review stage rather than on final fines, so it should not be applied to the 126,728 directly, but the direction is clear: even at a fraction of that rate, thousands of these fines would be wrong.
Here is the real imbalance. The detection side has been industrialised: cameras that never blink, scanning millions of cars, issuing fines at a scale no police force ever could. The dispute side has not kept up, and it is unforgiving of a weak case. Fire off a vague "it wasn't like that" note and you will most likely get a standard, copy-paste "no", as if no one truly looked. From there the only avenue left is to elect to go to court: months of waiting, real stress and cost. Cases that could have been resolved with one honest look end up funnelled into a courtroom that never needed to see them.
That is exactly why your review is worth getting right. In practice it is your one real shot to put things straight before that court path, so it has to be your best foot forward: the specific grounds that actually apply, the evidence that backs them, and the case framed the way the reviewing officer assesses it, not a hopeful sentence a reviewer can wave away in seconds. The camera is wrong more than one time in seven, and a single image often cannot prove what it claims. Make that one shot count and you may never need a courtroom at all.
So don't just pay. Ask to see the image first. Then, if it caught a moment that was not the offence, put a proper case forward, built on the kind of grounds a reviewer cannot easily dismiss. Common grounds include a belt that was worn but blended into clothing or sat behind your arm, a fine that relates to a passenger rather than the driver, a recognised medical exemption, the wrong vehicle, or an image simply too unclear to establish the offence.
Asking for a review is free and does not increase your fine, and the image, wrong more than one time in seven, often cannot prove what the camera claims. Fight My Fine is one of a new wave of tools Australians are using to do exactly this: it gives you a free, honest read on whether your grounds are strong, and if they are, drafts the review letter around what the reviewing officer actually looks for. It is not a law firm, and if your case is weak it will tell you so before you spend a cent.
In a single year, NSW issued more than 126,728 AI seatbelt-camera fines, a 2,357% jump on the previous year. Each carries about $400 and three demerit points.
Because 88% of the fines were for a belt worn incorrectly rather than not worn at all. Incorrect use, a belt behind an arm, low on the body, or slipped for a moment, is the hardest thing for a camera to judge from a single frozen frame, and it is where genuine mistakes cluster.
Yes. You can ask the issuing authority for a free review, and it does not increase the fine. The case rests on a single image, which often cannot prove the belt was not worn. Common grounds include a belt that was worn but blended into clothing or sat behind your arm, a passenger rather than the driver, a recognised medical exemption, the wrong vehicle, or an image too unclear to establish the offence.
Nobody can say for certain, because most people simply pay, so the errors are never counted. The clearest proxy is Western Australia, where around 60% of the AI seatbelt fines drivers challenged were overturned. Testing has also put the seatbelt AI false-positive rate at 15.28%, more than one in seven.
Fined because a seatbelt looked off in a camera image? A free, printable one-pager: what to check on the image, the grounds that work, the evidence to gather, and how to lodge a free review. Pop in your email and it's yours.
Sources: NRMA “Buckle up right” report (September 2026), citing Revenue NSW: 126,728 camera-detected seatbelt fines in a year, 88% for incorrect use, down 12.6% the following year; The Guardian via Redfern Legal Centre (revenue up ~1,400%, ~140 million cars scanned); The Conversation (are the cameras fair?). WA overturn rate and the 15.28% seatbelt false-positive figure from earlier AI-camera reporting.
Fight My Fine is a self-help tool, not a law firm, and this article is general information, not legal advice. You are the author and sender of every letter, and the issuing authority makes the final decision on any review. For serious matters or court, speak with a qualified lawyer or a free service such as your state's Legal Aid or community legal centre.