AI Road Safety Cameras Spot 19,000 Violations in Two Weeks: What Drivers Need to Know

Imagine driving along a perfectly ordinary road, one hand on the wheel, the other thumbing a message you "just need to send". No officer on the verge, no flashing lights, no marked van. Yet a camera has already captured your hand, your phone and your face, and software has decided that it looks like an offence.

That is the picture painted by a recent police trial of AI-based road-safety technology, reported by The Mirror. In just two weeks, the system identified 19,000 violations, including drivers using phones and passengers travelling without appropriate restraints. It is a big number, and it raises bigger questions about how we will be policed in future.

What happened

According to the report, the trial used artificial intelligence to scan passing vehicles for behaviour associated with serious road-safety risks. The headline findings were around 19,000 potential violations in a fortnight, with two categories singled out: drivers apparently using handheld phones, and people in vehicles not wearing, or not properly using, seat belts or child restraints.

The summary I have does not say which force ran the trial, how many cameras were used or how many alerts led to action. It also does not say how many detections were confirmed by a human reviewer. Those gaps matter, so treat the headline figure with some care.

If the 19,000 were spread evenly across the fortnight, that would be more than 1,300 a day. We cannot tell from the headline whether it came from one site or many, or whether every detection would survive checking. "Violations identified" is not the same as "fines issued" or "convictions secured". Detection is the first step in a longer process.

How this kind of technology generally works

Systems of this type, which have been trialled by a number of UK police forces and road authorities in recent years, usually combine high-resolution cameras with image-recognition software. The camera captures the vehicle, often through the windscreen and side windows. The software looks for patterns such as a hand holding a device near the face or lap, or a person whose torso does not show a diagonal belt.

Most schemes of this kind are described as "AI-assisted" rather than fully automated. The software flags candidate images, and a trained person then looks at them before any action is taken. I cannot say whether that was the process in this trial, but it is the usual model and the one that makes legal sense. Software is good at spotting "something that looks like a phone". It is less good at telling a phone from a sandwich, a sat-nav mount or a hand scratching a chin.

Why it matters

Phones and belts remain stubborn problems

Mobile phone use and seat belt non-use are not new problems. Both are well understood, widely publicised and still happening. For the police, they are hard to catch at scale. An officer on a bridge or in an unmarked vehicle can only watch so many cars. A camera that never gets tired, and can look into every vehicle in a lane, changes that equation.

Phone use and unrestrained passengers are also behaviours drivers tend to underestimate. A driver might think "I was only looking at the map", or "it is only a short trip to school". The law, as we will see, draws very few exceptions for either.

The scale of detection changes the odds

The deterrent effect of enforcement depends on how likely drivers think they are to be caught. Roadside patrols made it a low risk. Cameras that look into cars make it a higher one. Whether that changes behaviour is for the evidence to show, and I will not claim results the trial has not reported. But it is reasonable to expect more forces to ask whether the technology is worth adopting.

Privacy and proportionality

These cameras capture images of people inside private vehicles, including passengers and children who have done nothing wrong. That raises questions under data protection law (the UK GDPR and the Data Protection Act 2018) and the Surveillance Camera Code of Practice. Reasonable questions include how long images of compliant drivers are kept, who sees them and how accuracy is audited. Our earlier piece on AI road safety cameras: safer roads or surveillance explores that debate in more detail.

The legal angle

Mobile phones

Using a handheld phone while driving is an offence under regulation 110 of the Road Vehicles (Construction and Use) Regulations 1986. Since March 2022 the offence has covered holding and using a device for any purpose, not just calls and texts. Scrolling a playlist, taking a photo or checking a map in your hand can all count.

The fixed penalty is £200 and six penalty points. Drivers who passed their test less than two years ago can lose their licence on reaching six points under the Road Traffic (New Drivers) Act 1995. If the case goes to court rather than being dealt with by fixed penalty, the maximum fine is higher, and a ban is possible. Our guide to UK mobile phone driving laws and hands-free rules covers the detail, and why drivers can be fined at lights explains the stationary-traffic trap.

There are narrow exceptions. Calling 999 or 112 in a genuine emergency when it is unsafe or impracticable to stop is permitted. Using a phone to make a contactless payment is allowed when the vehicle is stationary and the goods or services are to be received at the same location, such as a drive-through. Hands-free use is not automatically lawful either. The police can still prosecute for careless or dangerous driving, or failing to have proper control, if you are distracted.

Seat belts and child restraints

Seat belt rules sit mainly in the Motor Vehicles (Wearing of Seat Belts) Regulations 1993 and related provisions. In broad terms:

• Drivers and passengers must wear a seat belt where one is fitted, subject to limited medical and other exemptions.
• A driver is responsible for making sure children under 14 are properly restrained.
• Children generally need an appropriate child seat or booster until they reach 135cm in height or their 12th birthday, whichever comes first.
• Passengers aged 14 and over are legally responsible for their own belts.

A fixed penalty is £100, rising to a maximum of £500 if the case goes to court. Seat belt offences do not usually carry penalty points, but the financial and safety consequences are serious enough. The wording of the summary, "without appropriate restraints", suggests the trial caught more than simply unfastened belts. Wrong-sized child seats and children sharing a belt are the sorts of things such cameras can potentially pick up, though I cannot confirm what this trial detected.

What happens after a detection

I cannot tell you exactly how forces in this trial processed their detections. In general, for road-traffic offences caught on camera, someone reviews the evidence, decides whether to take action, and then offers a fixed penalty, a driver-awareness course in some cases, or court proceedings. You can usually decline a fixed penalty and have the matter heard in court, but a court can impose a higher penalty if it finds you guilty. If you receive a notice, the official GOV.UK guidance on motoring fixed penalties and, where the stakes are high, a qualified road-traffic solicitor are the right places to turn.

What drivers should know

This is general information, not advice on any particular situation. With that in mind, here are some practical points.

1. Treat "hands-free" as a design brief. If a phone has to be handled while driving, it is probably the wrong setup. A properly secured cradle, voice commands and settings that silence notifications on the move all reduce the temptation. Even a mounted phone can attract attention if you are poking at it while moving.

2. Do not assume cameras are only at the roadside. Detection technology can be fixed, portable or vehicle-mounted. A lack of a visible patrol does not mean a lack of enforcement.

3. Check restraints before you move off. It takes seconds to confirm every passenger is belted and every child seat is the right type, correctly fitted and suited to the child's height and weight. Rental cars and lifts for other people's children are common slip-ups, so make sure you know your responsibilities before you set off.

4. Remember that you may be the one responsible for others. If you are driving children under 14, their restraint is your legal responsibility.

5. Keep what you can. If you receive a notice that you believe is wrong, such as an object mistaken for a phone, preserve any relevant evidence. That might include dashcam footage, the trip's purpose or details of any medical exemption. Respond within the deadline stated on the notice. Ignoring it tends to make things worse.

6. Know that images can be challenged. AI is not infallible. A notice relies on evidence, and drivers have the right to ask to see it and contest it. Whether a challenge is worthwhile depends on the facts, so for anything beyond a simple case, take proper legal advice.

7. Pay attention to the points. Six points for a phone offence is a serious matter for anyone already carrying points, and potentially licence-ending for new drivers. Insurance costs can rise too.

Looking ahead

This trial is one of a growing number of experiments with AI in roads policing. It points towards a future in which behaviour inside the car, not just speed or position on the road, is routinely monitored. We have already seen rapid growth in automated enforcement of speeding, and we covered the scale of that trend in 1m AI speed cameras: what UK drivers need to know.

Several questions will shape what happens next:

• Accuracy and oversight. How often does the software get it wrong, and how are errors measured and corrected? Public confidence depends on transparent answers.
• Legal clarity. Will courts and tribunals accept AI-assisted images as sufficient evidence, and what disclosure will defendants be entitled to?
• Proportionality. Is it acceptable to scan every driver to catch a minority of offenders, and what safeguards will apply to the images of those who did nothing wrong?
• Outcomes. The ultimate test is whether detection leads to fewer collisions and injuries. A big detection number shows the technology can find offences. It does not by itself show roads become safer.

The sensible approach for drivers is not to fear the technology, but to assume that the old gap between "technically an offence" and "likely to be spotted" is closing. A phone in the hand or a loose belt was always a risk to you and everyone around you. The difference now is that the risk of being caught is rising too.

For anyone who does receive a notice linked to this type of detection, stay calm, read it carefully and check the official guidance. Where the penalty could affect your licence or livelihood, speak to a professional who knows the area.