London Live‑Facial‑Recognition Trial: £320k Cost, 500k Scans, Zero Arrests, One False Positive

Key takeaway

The six‑month live‑facial‑recognition (LFR) pilot at London’s busiest rail stations scanned roughly 500,000 faces, cost £320,786, generated only one false‑positive alert and resulted in no arrests directly attributable to the system.

Trial scope and expenditure

Duration & locations: February – July 2026, 18 deployments across major London rail hubs (e.g., King’s Cross, Liverpool Street).
Faces scanned: More than 500,000 individual facial captures.
Financial outlay: £320,786 for equipment hire and police staffing, as revealed by a Freedom‑of‑Information request obtained by Liberty Investigates.
Police time: Almost 100 hours of officer labour were spent monitoring the system.

"Equipment hire and police staffing for the 18 deployments across the trial cost £320,786 and led to only one alert on a watchlist, which turned out to be an incorrect identification." – The Guardian, 29 Sep 2026

Operational performance

Alerts: The trial produced a single alert that matched a name on a police watchlist.
Outcome of alert: The match was a false positive; the identified individual was not on the watchlist.
Arrests: No arrests were made as a direct result of an LFR alert. Police did make unrelated arrests for assault, theft, weapons possession, etc., but these are excluded from LFR performance metrics.
Watchlist management: BTP emphasised that the watchlist was deliberately limited and controlled, with data‑quality and safeguarding procedures in place.

Extension and subsequent activity

Trial extension: In August 2026 BTP announced a four‑month extension, adding London Underground stations to the deployment.
Post‑extension alerts: Since the extension, three alerts have been confirmed, each involving individuals who were already complying with sexual‑harm prevention orders or other court conditions.
Policy framing: Transport for London (TfL) promoted the technology as a tool to target people on police watchlists and to combat violence against women and girls.

Expert commentary

Fraser Sampson (former biometrics commissioner, now Facewatch director)

"We know the technology works, but using it in supermarkets to deter shoplifters … is very different from deploying it to catch people on a public transport network. Success in a shop means no such people coming in. Success for the police trying to catch people means people being caught. In that respect the trial doesn’t appear to have been very fruitful."
Human‑rights perspective: A parliamentary joint committee on human rights called live facial recognition a “particularly clear example of risk” and urged stronger legal safeguards.
Political reaction: Labour MP Bell Ribeiro‑Addy called for a suspension of rollout until robust safeguards and a legal framework are in place.

Community and Hacker‑News discussion highlights

Scale skepticism: Users noted that 500 k scans over six months averages only a few thousand faces per day, suggesting many cameras were offline or operating intermittently.

"16 deployments for 6 months at rail stations – they must have been off 99 % of the time to only scan 500k faces."
ROI questioning: Commenters asked whether the pilot is a genuine crime‑fighting effort or a “giant fishing expedition”.
Watchlist size: One comment clarified that the watchlist contained roughly 450 individuals, limiting the system’s detection potential.
Potential misuse: Some users warned that surveillance could expand to tax enforcement, fare evasion, or broader civil‑liberties erosion.
Effectiveness doubt: A sarcastic remark highlighted the paradox of a system that reports success only when no one is caught: "My magical face detection software worked because … none of the undesirables came in!"

Broader implications for AI‑enabled policing

Cost‑benefit mismatch: £320k spent for a single false alert raises questions about the economic efficiency of large‑scale LFR deployments.
Data‑privacy concerns: Live facial scans process special‑category personal data; proportionality, necessity, and robust oversight are legally required under UK GDPR and the ICO’s guidance.
Bias and watchlist composition: Limiting watchlists to a few hundred names may reduce bias accusations but also curtails the system’s practical utility.
Transparency and accountability: Freedom‑of‑Information disclosures were essential to reveal the trial’s true performance; without them, public debate would lack factual grounding.
Future policy direction: The extension to Underground stations and the push to address violence against women suggest political momentum, yet parliamentary and civil‑society criticism may force stricter oversight.

Conclusion

The London rail‑station LFR trial demonstrates that, at current scale and cost, live facial‑recognition delivers negligible policing value while raising significant privacy and accountability issues. Policymakers must weigh the £320k expense and the system’s limited detection capability against the broader societal costs of pervasive surveillance before expanding the technology further.