AI Cameras See NSW Seatbelt Fines Jump By 1400 Percent

Imagine receiving over $1,200 in fines in a single week for a mistake you didn't even know you were making. For a Sydney driver named Isabel, this was a reality after she was flagged three times for wearing her seatbelt incorrectly.

“I was like, there’s been some kind of malfunction. This is impossible. I always wear my seatbelt,” she recalled. The photos from the new detection cameras revealed the issue: to avoid neck irritation from the strap, she had tucked it under her arm. Isabel is just one of tens of thousands of drivers in New South Wales caught by a significant expansion in traffic surveillance technology.

The Staggering Financial Impact

The introduction of AI-powered cameras has led to a dramatic surge in seatbelt-related penalties. In the 2024-25 financial year, the first full year of the technology's operation, revenue from seatbelt fines skyrocketed by an astonishing 1,400%. The state collected nearly $59 million from 132,698 penalties, a massive increase from the $3.7 million raised from 9,529 fines in the previous year. The government states that all revenue generated is reinvested directly into road safety programs.

The Common Mistake Catching Drivers Out

While the AI cameras scanned over 140 million vehicles, the vast majority of drivers were found to be following the rules. Only a tiny fraction, 0.09%, were flagged for a seatbelt offence. However, the data reveals a surprising trend: 88% of these offenders were not completely unbuckled. Instead, they were penalized for wearing their seatbelts incorrectly, just like Isabel.

A Tale of Two Fines

The automated nature of the fines has led to mixed outcomes for drivers. After receiving her first penalty notice, Isabel was fined twice more before she even had a chance to correct her behavior. She successfully requested a review and had the last two fines waived. “I went to Supercheap Auto and I got my bejewelled seatbelt cover, so I can wear it properly and it doesn’t irritate my neck,” she said. “Lesson learned.”

Kris Savic had a less successful experience. He was also caught wearing his seatbelt under his arm, but did so to avoid irritating stitches from a recent chest surgery. He explained this in his application for review but was denied because he had not provided a medical certificate. “I get now it’s wrong, but also it’s different getting caught by a police officer and being able to explain and show I have stitches, it might have helped me get a warning,” he reflected.

Do The Cameras Actually Make Roads Safer?

The primary goal of the cameras is to reduce road fatalities, as the government notes that 15% of deaths involved people not using seatbelts. However, despite the new technology, road fatalities in NSW are currently higher than at the same time last year, with 251 deaths compared to 208.

This has sparked a debate about the technology's effectiveness. Peter Khoury of the NRMA believes that while technology has its benefits, “the tip of the spear always has to be police.” He emphasized that visible highway patrols are the most effective way to change bad behavior and also raised concerns about privacy. Camilla Pandolfini of the Redfern Legal Centre worries about the impact of automated fines on disadvantaged people, who may have legitimate reasons for an offence but lack the resources or knowledge to challenge a penalty.

The Official Government Stance

NSW Roads Minister Jenny Aitchison defended the program, highlighting that the cameras can check vastly more vehicles than police officers—effectively monitoring every car on NSW roads about 20 times a year. She stated that over 99% of drivers are compliant and that an initial spike in fines was expected.

“Whenever enforcement is increased... it is to be expected that more penalties will be issued initially and then decrease over time as people get the message,” she said. According to the government's data, this is already happening, with a 60% drop in offences in July 2025 compared to when the cameras were first introduced a year earlier.