Police told not to close investigations until they have used facial recognition

Telegraph

01-04-2025

Police have been told to use facial recognition searches in every criminal investigation, The Telegraph can reveal.
The independent police inspectorate has urged forces to 'fully exploit' the technology after finding that some were using it more than others.
It comes as a Telegraph investigation reveals the true scale of police use of facial recognition, with forces conducting searches on the public every two minutes.
Officers are encouraged to obtain pictures of their targets – including witnesses and victims – from social media, doorbell footage and CCTV, and search them against the vast police national database (PND).
Police could be able to check driving licence photographs in future under plans to give police access to DVLA information, although the Home Office says it is not changing the law for that purpose.
Facial recognition technology is not subject to national guidance from either the Home Office or the College of Policing, which provides advice to police on conducting investigations.
The technology, introduced to catch serious and violent offenders, is now most often used for low-level investigations.
The reliance on digital technology has increased as police forces across the country cut officer numbers to reduce costs.
His Majesty's Inspectorate of Constabulary, Fire and Rescue Services (HMIFRS) praised those forces that carried out the most searches in a recent report, and recommended that no investigation should be closed until all images had been cross-checked against available databases.
The inspectors' most recent report said that by June, forces should introduce a new rule 'stating that when an image exists, investigators should search it against the PND and any other relevant databases before their force closes an investigation'.
Forces can adopt the recommendations at their own discretion.
Privacy concerns
The unprecedented use of facial recognition will be welcomed by some as a way of reducing time-consuming investigations, but privacy concerns have prompted MPs, regulators and civil liberties campaigners to urge the Government to impose new rules regulating its use by police.
The Telegraph found that one force, Essex Police, used the technology for 16 investigations into what was later determined 'non-crime'.
The force said it had used facial recognition for investigations that concluded that no crime had been committed.
Essex Police was criticised last year over its investigation into Allison Pearson, a Telegraph columnist, for remarks she posted online. The investigation was later dropped.
Last year officers ran more than a quarter of a million 'retrospective' facial recognition searches in the UK, including 30,000 by the Metropolitan Police alone – more than ten times the Met's figure for 2019.
Retrospective facial recognition can use images obtained from a range of sources, including CCTV, mobile phone footage, dashcam or doorbell cameras or social media.
Police obtain the images during their investigations and then run them against a vast police database to look for a possible match.
They can then approach the individual, without revealing that their image has been searched to find a biometric 'match'.
Home Office officials describe facial recognition as a 'key tool' for the police to identify suspects more quickly and accurately.
A search can cut the length of an 'identification' from around two weeks to just minutes.
The main resource available to forces is the PND, which collates records from 55 different agencies and holds 6.2 billion searchable records as well as millions of photographs and images.
Police are already allowed to search the passport database, but are required to ask for permission from the Government each time.
While live facial recognition cameras on British high streets have generated controversy, retrospective searches are far more commonplace and useful to police.
Some forces, including the Met, have bought private facial recognition software from the US to run more accurate searches.
Although facial recognition was first billed as a tool to catch serious offenders, including murderers and terrorists, data obtained under freedom of information laws by The Telegraph and Big Brother Watch shows it is now used to track anti-social behaviour, fly-tipping and mail theft.
Two forces – Humberside and Dyfed-Powys – said they had used the technology for 'level one' intelligence gathering by local policing. Most forces refused to reveal a detailed breakdown of how they had used the tool.
The Telegraph can also reveal that the Information Commissioner's Office (ICO) is investigating how police are using the software amid concerns that 'large amounts of sensitive personal data' have not been processed 'responsibly' and 'appropriately'.
An ICO spokesman said: 'People must be able to trust that technology is being used responsibly, and we are committed to ensuring appropriate use and improving public confidence about the safeguards surrounding its use.'
The controversial expansion of facial recognition has also prompted fresh privacy concerns from MPs and civil society groups.
The technology was backed by the previous Conservative government, but Chris Philp, the former policing minister, said he was now concerned about how it was being used.
'There are clear guidelines about how facial recognition should be deployed,' he said.
'Reports of misuse are deeply concerning and it should certainly never be used for non-crime investigations – in fact, non-crimes should not be investigated at all.
'Where used appropriately, new technology like this will be transformational for police forces suffering severe cuts under Labour.'
Unlike most other investigatory tools used by police, there are no national guidelines or policies that officers must follow.
The Home Office said that police were required to follow laws that govern personal data and privacy, including the Data Protection Act, Equality Act and Human Rights Act 1998.
But Big Brother Watch, the civil liberties campaign group, said the unfettered expansion of facial recognition was a 'scandal'.
'Police use of facial recognition scanning to identify people in a 'non-crime investigation' is deeply Orwellian and demonstrates how police use of this technology is dangerously off the leash,' said Madeleine Stone, a spokesman for the group.
'We urgently need a democratic, lawful approach to the role of retrospective facial recognition in the UK. Without this, police forces should not be using this intrusive technology at all.'
Sir Keir Starmer called for a 'wider deployment of facial recognition technology' after riots across the UK last year, when it was used to identify protesters.
There is currently no statutory oversight of the tools available to police because the Government's independent biometric and surveillance commissioner position is vacant.
Prof Fraser Sampson, a former holder of that role, warned in an article for The Telegraph that 'unlimited expansion' of facial recognition could see it deployed to monitor protests and industrial disputes where no crime had been committed.
'Surveillance is no longer about where the police put their cameras; it is about what the state does with the images from everyone's cameras,' he said.
'Uploaded selfies, pictures from dashcams, doorbells and social media posts generate incalculable numbers of facial images, all of which are accessible in perpetuity and can be searched against retrospectively.'
Calls for guidance to prevent police misuse
Grant Shapps, a former home and defence secretary, called for national guidance on facial recognition to prevent 'misuse' by police.
He said: 'Deploying [facial recognition] for minor non-crime incidents risks undermining public trust and eroding civil liberties.
'The right to privacy is a cornerstone of our democracy, and we must never allow convenience or expediency to trump fundamental freedoms. I would support clear national guidelines to prevent the misuse of this powerful technology.'
A Home Office spokesman said: 'Facial recognition is an important tool in modern policing that can identify offenders more quickly and accurately, with many serious criminals having been brought to justice through its use.
'All police forces using this technology are required to comply with existing legislation that protects people's right to privacy, but we will continue to work with experts in this field, and other interested stakeholders, to ensure that appropriate safeguards are always in place.
'We will set out our plans for the future use of facial recognition technology, alongside broader policing reforms, in the coming months.'
A National Police Chiefs' Council spokesperson said: 'We welcome the recommendation from HMICFRS.
'Retrospective facial recognition is an extremely accurate and invaluable tool that has been used by police for a number of years to great effect.
'Where there are reasonable lines of enquiry and a clear policing purpose, we encourage officers to use every tool at their disposal to assist in their investigations, including retrospective facial recognition.'
Why we should care about retrospective facial recognition
By Fraser Sampson
When I was preparing my annual report to parliament as Biometrics and Surveillance Camera Commissioner in 2022, an official said to me 'why do we care about retrospective facial recognition? Everyone's concerned about live'.
There are many reasons why we should care, and the freedom of information requests published this week reveal several.
The first reason is that police use of any facial recognition technology (FRT) – retrospective, live or predictive – is an extension of biometrics and that demands close and consistent attention. In all its uses of intrusive biometric technology, the state must be accountable.
A second reason is that the placement of live facial recognition cameras has dominated the public space surveillance debate, but surveillance is no longer about where the police put their cameras; it is about what the state does with the images from everyone's cameras. Uploaded selfies, pictures from dashcams, doorbells and social media posts generate incalculable numbers of facial images, all of which are accessible in perpetuity and can be searched against retrospectively. When it needed a human to do it, comparing even a handful of images was a resourcing decision, balancing effort and expense against likely outcome. With AI-enabled technology, mass matching has become cheap, quick and easy, and the discretion has evaporated. The old adage about looking for needles in haystacks will make no sense to the next surveillance generation. The search is a mouse click and the principal risk – to the police and the public – lies in not doing it.
There are further reasons why we should care about retrospective recognition. Photographs of convicted people are an obvious source of images for matching, but the state has other large collections of our faces. The UK driver's licence and passport databases have millions of high-quality images that can be used for retrospective matching, but we submitted them because we wanted to drive on a road or leave the country, rather than for a criminal justice purpose. More interconnected state databases exist and new ones will emerge, sometimes shared with other countries.
Another reason to care is that, without the signage, visible cameras and auto-deletion of live FRT, retrospective matching is invisible to the citizen. While the UK's regulation of covert surveillance is the gold standard, retrospective recognition takes place outside any formal framework, which seems at odds with the safeguards for surveillance generally.
Is live FRT really the conjuror's left hand, distracting our attention while intrusive retrospective matching takes place on an industrial scale out of view? Opponents of the technology might suggest so: rebutting their claims will rely on police chiefs providing the data needed to allow proper challenge, as some have done.
But surely the lawful, proportionate and justified use of police biometrics is too important to be left to individual statutory requests? The police also need to be better at making their case.
Freedom of information responses from Police Scotland show that they do not even collect the data needed to assure themselves of its legitimate use. Increasingly, there will be cases where the public expect the police to use retrospective facial capabilities – identifying victims and offenders in child sexual abuse image cases for example – and the police will find themselves explaining any decisions not to use the technology. That will also need reliable and consistent performance data.
I have heard some surprisingly poor arguments against the need for scrutiny of this technology from some policing leaders: 'if you've done nothing wrong, you've nothing to worry about' (a stunningly misconceived observation), 'it's just photography' (presumably in the same way that DNA profiling is 'just chemistry'). This simply reinforces the need for greater prescription.
Unlimited expansion of the technology also provides a compelling reason to care about it. Finding suspects is one thing but the technology is clearly being used in non-criminal incidents, raising questions around its boundaries and scope. How will it be used for monitoring public gatherings, attendance at protests, industrial disputes or during pandemic lockdowns? What contribution will it make to predictive policing? These are not hypothetical questions: when the College of Policing says the technology can be legitimately used to find 'potential witnesses', its reach is unconstrained – I cannot think of a living individual to whom that definition would not apply right now.
Policing is about connecting the dots and, as Steve Jobs said, you can only do that by looking backwards. That is precisely what retrospective facial recognition does, making it an indispensable crime-fighting tool. Enabling the police to use it with the trust and support of the public is perhaps the overriding reason why we should care about retrospective facial recognition.