Council to introduce facial recognition CCTV (www.bbc.co.uk)

Hammersmith and Fulham Council has approved more than £3m to expand its already dense CCTV network — the borough has over 2,000 cameras — by adding live facial recognition (LFR) at crime “hotspots,” AI tools on 500 cameras, and drones as an enforcement aid (subject to CAA approval). Plans call for two LFR cameras at 10 locations to match faces in real time against a defined police database, plus Retrospective Facial Recognition (RFR) capability to automatically search footage across the network to trace suspects’ routes. Introduction of LFR is contingent on police support; the Met currently runs a single static LFR pilot in Croydon with analysis due this autumn. For the AI/ML community this is a notable operational deployment of biometric systems at scale, raising technical and governance implications. Key technical points: real-time matching, automated vehicle tracking, and large-scale video indexing for RFR require robust models, low-latency infrastructure, and substantial labelled datasets; they also amplify risks of false positives, demographic bias, adversarial attacks and mission creep (e.g., non-crime enforcement uses). Civil-liberties groups warn of unprecedented mass surveillance and legal uncertainty — there’s no clear framework or safeguards yet — making this a live case study in the social, legal and technical trade-offs of public-facing AI-powered biometrics.