Digital speed camera at Ngauranga Gorge, Wellington. Photo: RNZ / Alexander Robertson
The company taking over the country's mobile speed cameras is running tests in the UK on how to use the AI-powered cameras to catch drunk drivers.
Australian company Acusensus is also testing its much more powerful cameras across the Tasman in Queensland to detect tailgating.
Then there are its trials in Western Australia to spot drivers breaking variable speed limits on motorways, like the type denoted on SH1 in Wellington which regularly display 70 or 80km/h limits, in place of 100km/h.
NZTA Waka Kotahi hastened to state to RNZ on Thursday that the mobile cameras here would only be used on speedsters.
"New Zealand legislation does not allow safety cameras to be used to enforce other areas of non-compliance (i.e. impaired driving or mobile phone use)", it said.
Yet its 2022 trials on Auckland motorways encompassed mobile phones and seatbelts offences, using Acusensus' 'heads-up' technology.
There was "strong penetration of states/counties who have trialled or deployed heads-up", the company said.
NZTA Waka Kotahi is most of the way through a years-long, costly process to take over the speed-and-safety camera network from police, leaving officers with just a few radar-pingers in patrol cars.
It has contracted the mobile part of the job to Acusensus for about $100 million over five years, offering drivers the reassurance that only the agency would decide on infringements and dole out penalties.
This is the first insertion of a private company into the speed camera system.
Acusensus made a gross profit of $A22m on revenue of $A49m last year, but said: "The revenue model is not linked to the number of offenders caught."
Government officials have reiterated that.
"Neither Acusensus or NZTA will receive any incentives or funds from tickets issued," NZTA said.
The five-year contract began in December, to deliver 80,000 hours of mobile camera use a year, 17,000 more than police recently achieved.
Acusensus has rapidly spread its operations to Europe, the US and UK, where what police have called "a world first" test against drunk driving began in December.
The trial was using AI "to try and change the way we police the roads forever", said a police video.
Waka Kotahi told RNZ on Thursday: "NZTA has not discussed this use of safety cameras with Acusensus."
Yet Acusensus describes its core mission as cutting down on distracted driving, not speed, and talks up changing driver behaviour.
"We can assess distraction," it told US media.
In its 2024 investor presentation, it said "over time, contracts are awarded, and then those contracts are typically expanded".
Its chief executive, Alex Jannink, told an Australia investor show in 2023: "There's a lot of artificial intelligence development to do all these new enforcement activities of, you 'know, mobile phone, seatbelt, and we are applying the 'smarts' of the team to trying to solve other challenges like impaired driving.
"How do we automatically work out who on the road network is impaired by drugs or alcohol?"
The company annual report said New South Wales had reported a sixfold drop in mobile phone offences since 2019 - the first place and time the firm's cameras came in - and a big reduction in road deaths.
NZTA could use such stats to argue to expand cameras into catching drunk drivers here.
Its $200,000 trial of Acusensus on Auckland highways in 2022 included detecting high levels of mobile phone use, and seatbelts not being used.
How it works
Acusensus' 'Harmony' system cameras are being deployed without all their capabilities being used, under the up-to-$20m a year contract with NZTA.
If all its capabilities were used for mobile phone detection - as in some Australian states - the cameras would take high-resolution pictures of every passing car.
A machine-learning artificial intelligence then combs through each picture for signs of a distracted driver on a cellphone. AI looks to see if both hands are on the steering wheel. If not, the computer then begins to look for objects which have the dimensions and appearance of a mobile phone or are being held in a manner consistent with a mobile phone.
Ones with no such signs are deleted.
If usage is detected, a package of still images and CCTV footage is sent, encrypted, to a Acusensus server on the Amazon Web Services Cloud database. This is in Australia, but once a database opens in New Zealand, the rules call for it to be sent there.
Sign up for Ngā Pitopito Kōrero, a daily newsletter curated by our editors and delivered straight to your inbox every weekday.
