Sussex Police Deploy AI Roadside Cameras Targeting Seatbelt and Phone Offences
TL;DR: Sussex Police switched on AI-enabled roadside cameras on 13 April, targeting drivers not wearing seatbelts and those using mobile phones. The system uses Acusensus cameras and follows a National Highways 2024 Sussex trial that detected 458 offences across seven days.
How the System Works
The trailer-mounted cameras capture images at two angles — a steep view through windscreens using an infrared flash for all-hours coverage, plus a shallower forward-facing angle that can work at high vehicle speeds. AI algorithms examine images in real time and flag candidate offences, which are then passed to human officers for verification before any enforcement action. The cameras can detect multiple offences at once.
Chief Constable Jo Shiner said the technology is meant to “enhance” rather than replace conventional roads policing, with the local rationale grounded in casualty data: during the last three years in Sussex, 82 people have been in collisions where a driver was using a mobile phone, and 214 have been injured in collisions involving no seatbelt.
Where This Sits in the UK Public-Sector AI Pattern
The deployment is the latest in a steady run of concrete UK public-sector AI rollouts rather than announcements. The Department for Transport has just published details of its Consultation Analysis Tool, which uses Gemini models on Vertex AI to extract themes from public consultations with human-in-the-loop review. The Ministry of Justice is widening an AI transcripts study for family courts. Each project differs in scope, but all follow a similar pattern: production deployment of a specific third-party model with a stated human-verification step.
The pattern is becoming the practical version of the UK AI Playbook — more so than the broader sovereign-datacentre announcements that have repeatedly slipped. Sussex Police, DfT and MoJ are all shipping AI into services that touch citizens directly, using commercial vendor stacks combined with procedural safeguards.
Looking Forward
The Sussex trial will run for several weeks before formal evaluation. Watch for two things: the rate at which AI-flagged images are confirmed by human reviewers (a proxy for false-positive cost), and whether the rollout pattern extends beyond Sussex and National Highways pilots to other forces’ business-as-usual enforcement. If the detection-to-enforcement ratio from the 2024 trial holds, expect similar deployments in forces with matching road-safety casualty profiles during 2026. For UK SMEs whose vehicle fleets cross Sussex regularly, the practical change is immediate: driver monitoring should assume camera coverage, not discretion.