This factsheet provides information on police use of facial recognition technology.
Types of facial recognition
The police use three types of facial recognition:
- Retrospective Facial Recognition (RFR)
- Live Facial Recognition (LFR)
- Operator Initiated Facial Recognition (OIFR)
Retrospective Facial Recognition
RFR is used after an event or incident as part of a criminal investigation. Images are typically supplied from CCTV, mobile phone footage, dashcam or doorbell footage or social media. These images are then compared against images of people taken on arrest to identify a suspect.
When there is a possible RFR match, a trained operator reviews the images to confirm it. An investigating officer also reviews the match to confirm accuracy.
The investigating officer will consider all of the evidence available and follow up all reasonable enquiries as in any normal investigation.
This is a key tool for the police to identify suspects more quickly and accurately. It can also help identify missing or deceased people.
A South Wales Police study found that without RFR identifications take around fourteen days, whereas with it they typically take minutes.
All police forces use the Police National Database facial search facility.
Real life examples of how forces use this technology to catch criminals and keep people safe include:
- Craig Walters was jailed for life in 2021 after attacking a woman he followed off a bus. He was arrested within 48 hours of the incident thanks to South Wales Police using facial recognition on images captured by CCTV, including on the bus.
- Images taken by a member of the public inside a Coventry nightclub where a murder had taken place were quickly matched on the police national database to a known individual. The victim’s blood was found on his clothing; he was charged and sentenced to life imprisonment.
- A two-year operation was undertaken to disrupt, deter and build intelligence of drug operations within communities in Cardiff. RFR identified those involved and generated an intelligence picture of drug dealing in the city. This led to the arrest of 69 drug dealers, with 64 being charged. 44 people are serving custodial sentences totalling 117 years.
- Multiple burglars have been caught thanks to doorbells with cameras, the modern equivalent of a thief leaving his fingerprint on a windowsill. A man was jailed for burgling the home of a vulnerable pensioner. Using RFR on a CCTV image from the pensioner’s home, officers identified the suspect, arrested him the following day, and he later pleaded guilty to burglary.
Live Facial Recognition
Live Facial Recognition (LFR) enables police to identify wanted people, a core part of policing. Every day, police officers are briefed with images of suspects to look out for. Over many years, officers have spotted individuals from these briefings and taken action to keep the public safe.
LFR does what the police have always done but much more quickly and with greater accuracy.
All deployments are targeted, intelligence-led, time-bound, and geographically limited. It lets forces place their effort where it is likely to have the greatest effect. Before a deployment, the police will inform the public where they intend to use the technology and where they can obtain more information on its use.
The technology uses live video footage of crowds passing a camera and compares their images to a specific list of people wanted by the police. The technology can precisely pick a face out of a dense crowd, something which would be impossible for an officer to do. It means the police can quickly and accurately identify wanted criminals and take them off the streets.
Following a possible LFR alert, it is always a police officer on the ground who will decide what action, if any, to take. As with normal investigations, the investigating officer must form reasonable grounds to suspect the person identified is responsible for the commission of an offence and that there is justification for an arrest. The standard procedures for investigation, evidence collection, arrest, charge, and prosecution are followed.
If the LFR system does not make a match with the watchlist, a person’s biometric data is deleted immediately and automatically. The watchlist is destroyed after each operation.
It is an operational decision for individual police forces whether, how and when to use the technology, in line with the College of Police Authorised Professional Practice. To date, South Wales Police (SWP), the Metropolitan Police Service (MPS) and Northamptonshire Police have used LFR.
Real life examples include:
- At the Arsenal v Tottenham north London derby on 24 September 2023, it led to three arrests, including a suspected sex offender.
- A wanted sex offender was sent back to jail after being identified at the Coronation of King Charles. An image of his face matched that of a wanted suspect. He was arrested and sent back to prison for breaching the terms of his release.
- Over two busy Friday nights in Soho in August 2023 the Metropolitan Police used it to help find high harm offenders. Across the two deployments there were six accurate alerts and no false alerts. It led to the police engaging with six people, five of whom were arrested including a man wanted for possession of a bladed article and a woman wanted for breach of bail in relation to robbery.
Operator Initiated Facial Recognition (OIFR)
OIFR is a mobile App that allows officers, after engaging with a person of interest, to photograph them and check their identity where they are not sure, without having to arrest them and taken them into custody.
It is at the early trial stage but has been showing positive results.
Frequently Asked Questions
Why is the Home Office encouraging the police to make more use of facial recognition technology?
- The Government is committed to making sure the police have the tools and technology they need to solve and prevent crimes, bring offenders to justice, and keep people safe.
- Technology such as facial recognition can help the police quickly and accurately identify those wanted for serious crimes, as well as missing or vulnerable people. It also frees up police time and resources, meaning more officers can be out on the beat, engaging with communities and carrying out complex investigations.
- The Home Office is working closely with industry and the National Police Chiefs’ Council to keep pace with the rapid developments and improvements in this technology and make sure forces can use it in an effective, fair, and proportionate way.
Didn’t the courts say that police use of live facial recognition was unlawful?
- No. The Court of Appeal in Bridges in 2020 found that there is a legal framework for police to use LFR.
- However, it also found that South Wales Police did not fully comply with privacy, data protection and equality laws during two of their LFR pilots.
- Since then, the police have addressed the Court’s findings:
- The College of Policing has issued Authorised Professional Practice guidance on live facial recognition, in particular setting out the circumstances in which the police can use it, and the categories of people they can look for.
- The National Physical Laboratory has independently tested the algorithms South Wales Police and the Metropolitan Police use, and found they were very accurate and there were no statistically significant differences in performance based on gender or ethnicity at the settings the police use.
What laws govern police use of facial recognition?
- There is a comprehensive legal framework in the UK, which means that the police can only use it for a policing purpose, where necessary, proportionate, and fair.
- The police have common law powers to prevent and detect crime. At the same time, they must comply with data protection, human rights, and equalities law.
- This means that all deployments must be for a policing purpose and be necessary, proportionate, and fair.
- They need to comply with any published policies and, in the case of Live Facial Recognition, follow the College of Policing guidance.
What safeguards are in place to make sure facial recognition is operated fairly and without bias?
- Its use is governed by data protection, equality, and human rights laws. It can only be used for a policing purpose, where necessary, proportionate, and fair.
- The College of Policing national guidance sets out when the police can use LFR, the categories of people they can look for, and the requirement to automatically delete the biometric data of anyone the system does not match to the watchlist.
- Following a possible LFR alert, it is always a police officer on the ground who will decide what action, if any, to take.
- When there is a possible RFR match, a trained operator reviews the images to confirm it. An investigating officer also reviews the match to confirm accuracy.
- The investigating officer will consider all of the evidence available and follow up all reasonable inquiries as in any normal investigation.
- As with normal investigations, the investigating officer must form reasonable grounds to suspect the person identified is responsible for the commission of an offence and that there is justification for an arrest.
- The standard procedures for investigation, evidence collection, arrest, charge, and prosecution are followed.
- Facial recognition technology will also never replace the need for human judgement, insight, and empathy. This is not automated decision making – police officers will always make the decisions about whether and how to use any suggested matches.
Is it true that facial recognition is not as accurate with female or non-white faces?
- Accuracy depends on the algorithm used but is improving all the time.
- The National Physical Laboratory (NPL) tested the algorithms South Wales Police (SWP) and the Metropolitan Police Service (MPS) have been using.
- At the settings they use the NPL found that for:
- LFR - there were no statistically significant differences in performance based on gender or ethnicity in the way they use it. In practice, at the time of publication, there have been no false alerts this year.
- RFR and OIFR - 100% accuracy in identifying a correct match with no false matches.