I beg to move amendment 33, in clause 21, page 17, line 33, after “subsection” insert

“, and may make consequential amendments of this section”.

This amendment will ensure that, in the event that the list of bodies to which the new provisions apply is amended by the regulations, any necessary consequential amendments may also be made by the regulations.

With this it will be convenient to discuss the following:

Government amendments 34 and 35.

Amendment 60, in clause 21, page 19, line 11, at end insert—

“(4) The Secretary of State must lay an annual report before Parliament providing information on the use of the powers introduced by this section.”

This amendment would require the Secretary of State to publish a report on the police’s use of the new powers giving them access to driver license records.

Clause stand part.

Clause 21 will facilitate the ability of police and other law enforcement officers to access driver information from the Driver and Vehicle Licensing Agency automatically at the point of need for all policing and law enforcement purposes. The clause contains provisions for regulations and a code of practice to enable changes to be made for the purposes of access to driver data.

Police and other law enforcement officers frequently need to deal with unpredictable circumstances while operating under time pressure, and the police have provided compelling operational examples in which they could deal better with threats and harms faced by the public if they had direct and automatic access to DVLA driving licence data. That is why the Bill contains this clause.

Currently, automatic access to the DVLA driving licence data is limited to matters concerned with enforcing road traffic offences and Road Traffic Act 1988 purposes, which means that the police do not have automatic access to driving licence data for other purposes, such as investigating wider types of crime. They do have manual access to the data, but that is quite laborious and takes a long time. The clause builds on an existing regulation-making power and requires the Secretary of State to make regulations specifying the purposes for which, and the circumstances in which, the police can automatically access the driving licence database, so that all policing and law enforcement purposes are included.

The clause essentially expands an existing manual process and makes it automatic. At the moment, the automatic access applies only to road traffic matters. Given that we hold this data, I am sure that members of the Committee would expect the police to be able to access it in pursuit of criminals. Through the clause, we are delivering on our commitment to support the police by providing them with the tools they need. Members will recall that in the oral evidence sessions back in December, we heard support for the provisions from the National Crime Agency, the College of Policing and His Majesty’s chief inspector of constabulary and fire and rescue services, Andy Cooke.

We are also taking the opportunity to make the legislation simpler and more transparent. Under the current system, some law enforcement staff have access to the DVLA driving licence database and others do not. The clause allows all police and law enforcement organisations listed in proposed new section 71(4) of the Criminal Justice and Court Services Act 2000 to access DVLA driver information. It also changes the current legislative quirk whereby police staff are defined by law as police constables solely for the purpose of DVLA driver data, which is rather anomalous. It will define those permitted to access DVLA driver data as “authorised persons”, and regulations will set out the process for those individuals to become authorised.

We are conscious that police access to such personal data needs to be reasonable and proportionate, as set out in existing data protection legislation. That is why the clause will bolster the governance arrangements in this area by making them more transparent and straightforward, in preparation for the regulations that will broaden the access regime. In particular, the clause makes it clear which bodies can access DVLA driver data, which is necessary as the current legislation could be improved upon. There is a mechanism to amend the list of organisations that have access, while requiring the chief officer to be responsible for ensuring that appropriate training and oversight is provided by their organisation. That obviously includes the delivery of their data governance obligations.

Government amendment 33 ensures that if regulations amend the list of law enforcement bodies in subsection (4) of proposed new section 71 of the 2000 Act, such regulations may make consequential amendments to section 71—for example, to the definition of “chief officer”. Amendments 34 and 35 clarify the meaning of the term “chief officer” for each of the listed law enforcement agencies.

With your permission, Ms Bardell, I will speak to amendment 60 once I have heard the shadow Minister’s comments so I can respond to them. I hope I have set out why it is important that we put on a sounder and clearer footing the organisations and individuals who can access this data, and lay the groundwork to make automatic access possible for law enforcement purposes. I am sure we can all get behind that.

I want to speak to amendment 60 in my name. I do not know whether the Official Report can capture this, but my feeling is, “Hmm!” There are things I hoped the Minister would say, not least because it would have reduced my contribution by half, but he did not, so I will have to see whether I can get him to say them.

This is a very important clause. It is a long-standing, accepted principle that under certain circumstances police officers can use driver information held by the Driver and Vehicle Licensing Agency. At the moment, it is for purposes related to the enforcement of road traffic offences. It is a well-established and uncontroversial principle that officers can talk to the DVLA and get information. We have no quibble with the clause seeking to make that a smoother process, and we have no problem with subsection (4) of proposed new section 71 of the 2000 Act, which broadens and clarifies who can ask for that information, but the clause potentially contains a very broad extension of powers.

Paragraph 370 of the explanatory notes says:

“This clause replaces existing section 71 of the Criminal Justice and Court Services Act 2000 with a new section 71, and inserts a new section 71A into that Act. It provides for police and law enforcement bodies to have access to driving licence data held by the Driver and Vehicle Licensing Agency (“DVLA”) and provides powers for the Secretary of State to make regulations concerning the purposes for which such data may be used and any conditions which need to be met by any person accessing the data.”

That is a really significant change. The police had access to DVLA information in relation to a very narrow set of offences, but the clause is essentially saying that the information is fair play for enforcement agencies, subject to whatever regulations a Secretary of State may set in the future. That is an egregious use of Henry VIII powers. If the Government have a regime in mind, they should say what it is. Instead, we are possibly giving over the entire DVLA database to the Secretary of State, and it is currently not clear what it will be used for.

First, for clarity, I should say that the purpose for which the information might be made automatically accessible is circumscribed by the words at the very end of the newly created section 71(2), which says the that Secretary of State

“may prescribe only such purposes and circumstances as are related to policing or law enforcement.”

It is not for any purpose; it has to be related to policing and law enforcement.

Secondly, I pointed out that the police already have the ability to access the data manually for police and law enforcement purposes. The clause simply paves the way for them to do what they currently do manually and instead do it automatically, provided that it is for a police or law enforcement purpose.

I hope those two points give some reassurance that the clause is not some gigantic land grab. It simply makes automatic what is currently manual and it is circumscribed in the way that I described.

The second point and the tone of the Minister’s contribution certainly give me a degree of comfort. The first point does not give me any comfort at all. The line that he quoted, which provides that the Secretary of State

“may prescribe only such purposes and circumstances as are related to policing or law enforcement”,

is a landing zone as wide as could possibly be needed to give police access to the DVLA database for all sorts of things. I will get on to that shortly. I would be delighted if the scenarios I am about to talk about are not what the Government intend, or even silly ones. I do not think that is the case, but I will be the gladdest person in the room if it is.

At the moment, I think the public would think there are sensible uses of the data, such as to double-check the last known addresses of a wanted person. Taken in extremis, however, the clause could mean that all the photos on the DVLA database become fair game for law enforcement. There are 41.2 million vehicles registered with the DVLA. We are talking about the data of tens of millions of people and, crucially, their photographs. The landing zone that I described under subsection (2) of proposed new section 71 means that someone’s photo could be used for a huge range of things.

Starting from where we think the public would be on this, the public expect and know that retrospective facial recognition is already used routinely. If a person commits a crime, they know that their custody photo will be held on file and that, if they are convicted, it could be held for six to 10 years, depending on the nature of the offence. It is a tale as old as the technology that the photo might subsequently be checked against footage relating to another crime and used for its detection. Similarly, if someone commits an act that means that their DNA is kept on file, that might be used in relation to unsolved crimes that happened in the past or to tackle offending in the future. All that is sensible.

On both fronts, we start in a sensible place, with the use of the DVLA database for cross-references to help detect crime, and the use of photos of offenders that lie on file to detect crime. The risk comes when we mix those two sensible principles.

I have listened carefully to my hon. Friend. Does he agree that there is a danger that we will get mission creep and, more importantly, that the public will see it as such? As time moves on, we will get another bit, another bit and another bit.

I absolutely agree, and that is exactly the point I am about to make. It is a legitimate anxiety. If that is wrong and I am somehow being conspiratorial, albeit it is not my nature, the Minister can tell me and that would give me a degree of comfort, but I think, and it has been expressed to me by interest groups, that there is a legitimate anxiety that the clause could mean that the police will access tens of millions of driver photos to solve crimes.

There is a problem with that on two fronts. First, we have the question whether that is a fair balance between individual liberty and collective safety. If that is what the Government want, we need to have that debate much more clearly before we get to these specifics. Secondly, certainly for the photos that are on file today, probably each one of us in the room, Members and beyond, is on that database, but none of us will have given our photograph with informed consent, knowing they were going to be used for the proposed purpose. That is a huge problem, not least because it is a kind of coercion. I would still give my photo on that basis, because the ability to drive is fundamental to my needs—certainly for work, never mind family life. But, again, it is a form of state coercion, rather than something that would necessarily be the right thing to do. Many people—many of my constituents, I know—would not consent on that basis. We live in an age of high-profile conspiracy theories and those who make an industry of them, and we should not give grist to the mill of yet another one.

rose—

I give way to the Minister; nothing would please me more than to be contradicted.

Existing DVLA data, including photographs —I have a driving licence; I am sure that the hon. Member has one and everybody on the Committee probably does as well—can already be searched by the police; or rather they can request information on a manual basis. That can happen already, to each and every one of those 40-odd million records, for law enforcement purposes. That has a statutory basis, via schedule 2 to the Data Protection Act 2018, which allows access to that kind of data for police and law enforcement purposes. The principle of police access to that data for law enforcement purposes is already there.

I pose a brief example: if somebody was wanted for an offence such as rape, and the police could not track them down and they were not on the custody image database—let us say that they had never been into custody—would the Committee want the database to be searched to see whether a recognition match could be obtained to find that person? That can happen at the moment, and I think that is right. I would like to check whether the shadow Minister agrees.

There is a significant difference between what the Minister talks about, which is using, on an individual basis, information that is available to cross-reference—perhaps to deal with known aliases or known addresses—and where this could go. I mean a broad-scale use of surveillance, seemingly without any guardrails, that would essentially allow tens of millions—I think it is in the 40-something millions—of photos to be put on essentially the same footing as photos from the custody suite.

I heard the Minister’s intervention, but I am increasingly concerned about the point I made about mission creep and proportionality. That is what we must bear in mind. Would my hon. Friend agree? This partly goes back to amendments 61, 58 and 59: let us be as precise and specific as we can—in fact, let us have some control rather than this mission creep, and always have human rights and the question of proportionality in the back of our minds.

“Proportionality” is the exact word here. It is about finding balance, as I say, between individual liberty and our collective safety. If we are saying that, in all instances where the police have a photo of a possible offender, we have complete comfort about their searching the entire DVLA database to try and identify that person—well, that is a very significant change. I would argue strongly that that is different from saying, “Well, we think we might know who it is; let’s apply for the driving licence of an individual.” I think that the public would see a big difference.

The issue becomes even more important given the pace of technological change; we have been talking about retrospective facial recognition technologies here, but live facial recognition technologies are developing at breakneck speed: cameras can scan huge crowds of people, check who they are and flag persons of concern. The Minister, I think, has a degree of enthusiasm for that technology; it is a plank of his plans around violence against retail workers, and shop theft more widely.

I recently visited South Wales police because I wanted to understand the use of this technology. With your indulgence, Ms Bardell, I will take this opportunity to put on the record my thanks to the police and crime commissioner for South Wales, Alun Michael, for his work. He is stepping down from public office at May’s elections after more than half a century in different posts. He has sat in the chair that I occupy; he has sat in the chair that the Minister occupies; he has sat in Cabinet, and he is one of the few PCCs to have served the entire 12 years of that post’s existence. His is a life of extraordinary public service to the UK and Wales, and to Cardiff, and we have been very lucky to have him.

As I was saying, South Wales police is very much in the vanguard of facial recognition technology. In a recent oral questions, the Minister graciously made reference to the work of South Wales police in this area, particularly the oversight that it is doing through the PCC’s office, which may well be a good model—in fact, it very much is. However, it is not without controversy, partly because it is in the vanguard and that sort of disruption is generating significant public interest.

There was quite a bit of public debate around one of the events in question: a Beyoncé concert at the Principality Stadium. That was a huge event. It had incredible hype; people really did everything to try to get tickets. Live facial recognition technology was used at that concert. The police advertised it, defined a public watch list, and said, “We are looking for anyone who is a known terrorist, in light of the Manchester bombing, and for paedophiles, in view of the likely attendees of the concert.” That probably finds the right balance: the public are informed and there is a defined watch list. There is a degree of common sense there.

As I said, what South Wales police have put in place, and what I think we are going to see across the country, is a model where we have civilian balance to the operational independence of the police and where the public’s representative, as we have in England and Wales in the form of police and crime commissioners, is able to know what is happening and to have public oversight. However, there is a real risk around how that sort of technology could be combined with the powers in this clause. The use of live facial recognition technologies could be combined with the entire DVLA photo database for the purposes of live blanket surveillance. I do not think that is what the Government want, but I do think that we have to hear that they do not. Relying on Henry VIII powers makes it even worse.

It is welcome that, according to proposed new section 71A of the Criminal Justice and Court Services Act 2000, a code of practice will be issued. That is really important, and speaks to my amendment 60, which is a pretty basic attempt to put some degree of parliamentary guardrail on to this new technology and its use in law enforcement. It states that once the Bill has received Royal Assent, a report on its usage must be published by the Secretary of State on an annual basis. I will be honest: I am not precious about the form that that should take, but I am precious about technology that has moved at great pace and with which there has been very little parliamentary engagement at all.

Stronger guardrails should be set. We have to be really cautious about other changes we might make, such as those in clause 21, that would mean that a significant database could be used for proactive surveillance of people who have never done anything other than apply for a provisional licence. That point could be held, but we would have to argue with the public and build public confidence around it. We would have to be shown the evidence for the effectiveness and the operational cases. I do not think the Government have done that.

My hope is that I am taking the issue far beyond what the Government intend. I will be very glad if that is the case. Part of the problem is that we are relying on regulations when we have no idea what their form will be. I stop at this point. I hope the Minister will be able to give the Committee some comfort.

I have listened extremely carefully to the shadow Minister and the Minister. I take issues of data protection and privacy very seriously, but I also believe that members of the public expect us to try to keep them safe, and expect the police to be able to use technology to try to keep them safe.

I listened to what the hon. Member for Nottingham North said about the facial recognition technology used in south Wales. Those same south Wales cameras were used last half-term in my constituency of Chelmsford and in Southend in Essex. As a result, a number of individuals were arrested, including somebody who had come into Chelmsford from outside and was wanted for a very high-profile sexual offence. What the Minister has been saying about being able to use this technology to arrest rapists is not hypothetical; it has happened, and it happened in my constituency in Chelmsford.

I am sorry to ask the right hon. Member to give way, because I have said so much—but I guess the point I was making about the Beyoncé concert was about having a defined watch list of people known and suspected of having done significant crimes. Such measures feel like a proportionate tactic to apprehend them, but is that enough for us to say that therefore everybody’s faces ought to be in play, even though the vast majority have never done anything wrong?

I was just coming to that point. I completely hear what the hon. Gentleman is saying about widening out the potential dataset of photographs that could be used. I suspect that most members of the public would say it is utterly reasonable to widen out the set of faces we could look at. For instance, we might have a picture of somebody who is a rapist and was never known before, but their photograph might have appeared on a wider database such as that from the DVLA, rather than on a known potential suspect list. If that helped to find really dangerous criminals, I suspect most members of the public would accept that.

I agree that there need to be very strong measures to make sure that we do not turn into a surveillance society, such as we have seen in some much more autocratic countries, where those pictures can then be used against innocent people. I believe the Minister is trying to say that the protection is there in the Bill. When these cameras have been used, there have been very strong rules about where they can be used. The cameras used in both Chelmsford and Southend were highly publicised, with signs and alerts that the technology was being used in that place on that day. Those sorts of measures are important.

I would be interested in a bit more clarity from the Minister about ensuring that we do not turn to inappropriate use, as we have seen in some autocratic countries, while also ensuring that we can help the public by making use of the best technology to identify those who have committed the most serious crimes.

I would like to speak to a couple of matters concerning the clause, and then talk about the read-across into facial recognition. I think there is some conflation of two different kinds of facial recognition—live and retrospective—only one of which is relevant here. I will explain how that works and how it might be relevant.

Let me speak first to the to the substance of the clause. First, just to be precise, this is not a Henry VIII power, which is a piece of secondary legislation that can amend primary legislation. This does not create a regulation-making power to amend primary legislation; it simply creates a list. It is not technically a Henry VIII power, but it does specify the circumstances, or the policing and enforcement purposes, for which data can be used.

Secondly, we are not passing clause 21 expressly to authorise the use of DVLA data for facial recognition or indeed for any particular law enforcement purpose; we are simply creating a regulation-making power. Regulations might be brought forward using this power to do any number of law enforcement activities, including facial recognition, but that would, of course, have to be debated at the time and subject to the affirmative procedure. This is enabling legislation; we are not actually passing the substance of any particular power today.

Since members of the Committee have raised questions about facial recognition, I would like to address them. There are essentially two different kinds of facial recognition—separate concepts—one of which is relevant and one of which is not. Let me start with the one that would not be relevant here: live facial recognition. Live facial recognition would never need to use images from the DVLA database, the passport database or anything else. Live facial recognition takes a watch list of people wanted for criminal offences or wanted by the court.

Such technology was run recently in Croydon—my borough—where I think there are about 12,000 people wanted for criminal offences. The camera was set up, as it was in Southend and Chelmsford, in Wales, for the coronation and at the Silverstone Grand Prix. Everyone who walked past those cameras in Croydon town centre got scanned. There were public signs up; the authorised professional practice says that public signs have to be put up, which I would have thought would mean criminals would not walk past the cameras—but they do.

Over a couple of afternoons in Croydon, several thousand people walked past the cameras, despite the fact that there was a sign up saying, “The police are using live facial recognition” and in the course of those two afternoons, there were 17 alerts and all were accurate. The police would stop each person, have a conversation, ask for ID and so on, and all 17 had been accurately identified. Those 17 people were all arrested because they were wanted for offences. One of them was a wanted rapist who would otherwise not have been caught. It was just good fortune that that wanted rapist had to walk past the camera and was apprehended. Were it not for that, that wanted rapist would still be wandering around. There was another person who had been wanted for grievous bodily harm since 2015—for eight years—and they were arrested as well. In Croydon, the scheme was run in the first couple of weeks of December.

Concerns have been raised previously about accuracy and disproportionality: does the technology pick on members of certain ethnic communities, as it were? There was litigation that preceded South Wales using it, back in about 2020. I join the shadow Minister in commending Alun Michael for his work on live facial recognition, where he has shown real leadership, and for his long career in public life—in Parliament as a Secretary of State, as Welsh First Minister, and as a police and crime commissioner for the last 12 years or so.

The litigation happened, and the court set out certain conditions that had to be met, one of which was around accuracy—I think the false positive rate had to be one in 1,000—and disproportionality. The algorithm was improved and tested by the National Physical Laboratory at 0.6, which is the setting that the police use. At that setting, it found that the false positive rate was, I think—I am speaking from memory—one in 6,000, which is six times more accurate than the court specified, and that there was no disproportionality based on either gender or ethnicity.

For the one-in-6,000 false positive, what would happen—this has not happened in Croydon, because only 17 people were stopped—is that the police would say, “Excuse me, sir or madam, can you just identify yourself?” If they turn out not to be the person the police think they are, obviously the police will say, “Sorry to disturb you—off you go.” But that is one in 6,000, and it did not happen in Croydon, because the police did not stop 6,000 people.

So that is live facial recognition. The police do not need access to a DVLA database or password database, or even the custody database; they just need a list of people who are wanted. In the examples in Chelmsford, Southend and Croydon, they caught people who were wanted for offences or who had absconded from court, who would otherwise still be wandering around. It was used at the coronation. Three camera points were set up and, similarly to the Welsh example, the watch list included known terrorists, fixated people—people who are obsessed by the monarchy—and also wanted offenders. I think three wanted people were caught on the coronation day, including a wanted sex offender who happened to wander past the camera. The technology has also been used in Soho to try and catch robbers who were stealing mobile phones and so on.

I hope that illustrates to the Committee how live facial recognition is useful. It comes under the auspices of the Supreme Court judgment in the Bridges case, and authorised professional practice is published by the College of Policing. I think the Information Commissioner’s Office has also looked at it. Live facial recognition, which does not need DVLA data anyway, has a good basis, and it is catching people who would otherwise not get caught.

We can all come up with our examples, whether they be of rapists or a missing person, for example—who would not want to find a missing person who might have been missing for years and who we are trying to find? That is not the issue. We have used DNA in the past, and we all think it is wonderful, but it is used within parameters. Yes, let us have CCTV and facial recognition, but they have to be used within a certain context.

Then there is body-worn video and the move towards using algorithms in the detection and prevention of crime —and, in some places, the deterrence of crime—and also the issues around AI. I do not think there is a problem with that at all, but it has to be set in a context. We should not take a piecemeal approach to the whole question of technology. We are dealing with a vast area, and this must be set in that wider context. Perhaps the Minister can go off and think this through, outside this Criminal Justice Bill, and set in process a review of the whole question of artificial intelligence and so on, as part of crime reduction, crime prevention and crime detection.

The hon. Gentleman is making a very reasonable point. In fact, I was going to come on to it later, but I will address it now. As for the current deployments of live facial recognition in places such as Wales, Croydon, Southend, Chelmsford, Soho and elsewhere, a framework was set out by the case of Bridges v. South Wales Police. There is authorised professional practice from the College, but that framework is there. The hon. Gentleman is, however, making the valid point that it would be useful to set out—outside of this Committee, as he said—a broader strategy, framework or explanation of how the technology can be used. That is not just the case for live facial recognition, but also for the retrospective recognition, which I will come on to in a minute. It is all set out, and we could perhaps debate and discuss it in the round.

The hon. Gentleman is making a reasonable point about something that I had in mind to do in the relatively near future. That is something that I am working on in the Home Office, and his suggestion has spurred me on to expedite that. We want to have this set out transparently and comprehensively so that everybody can look at it and know where they stand. It has actually been done in relation to live facial recognition, but it would be helpful to set it out in a broader context.

The point that my hon. Friend Peter Dowd made is far more eloquent than mine. The Minister is clearly very enthusiastic and learned about this issue. When he was talking about how live facial recognition does not rely on DVLA data or any photographic data at all, the Minister said that we just need a list. What does that rely on, then? It must rely on some sort of photographic data.

My second point—I will try to be slightly more eloquent and not sound like I am asking my kids to turn the telly on for me because I do not understand it—is about my kids. I do not mind my DVLA data being used; I will tick the box or whatever. It strikes me, however, that my son has a driving licence, and I feel slightly anxious about the way that large groups of young boys get treated like they are a rabble. There are all the brilliant examples of this, such as finding missing people or catching rapists—“Great. Wahey! Let’s lock them all up”—but I feel slightly less comfortable about my son’s photo being used, funnily enough, in what might be a broad-brush police and crime approach.

I thank the hon. Lady. On the first question, which was about how live facial recognition works, we first of all need a watchlist—the list of wanted people—and then the photographs. Maybe there will be someone who is wanted for rape, and we have that person’s photograph. Maybe it is someone the police have encountered before, but they just cannot find them. Maybe the police have gone to what they think is the person’s home address, and the person is not there, but they have their photograph. They have that data. Then, in the case of the Croydon deployment in early December, there were 12,000 photographs of people who were wanted for different offences—the police had those photographs.

Then, the faces of people walking down the road are scanned. Each time someone walks past, their face is scanned to see whether it matches. For the vast majority, there is no match, and the scan is instantaneously, immediately and automatically deleted. There is no need for any manual intervention; it gets deleted straight away, like this—although I am not sure how my clicking my fingers will be rendered in Hansard. If the hon. Lady or I had walked past, we would be scanned and, one hopes, there would not be a match. Then there would be instant, automatic deletion. If someone is scanned and their photograph is one of the 12,000 on the list, however, there is a flash—that was a double click, for the benefit of Hansard—and an alert is sounded. The officer will then go over and say, “Excuse me, sir”—all of the 17 who were stopped were men, by the way.

You shock me!

The officer will go over and say, “Excuse me, sir, what is your name?” When he has established that the match is accurate and that they are who the system thinks they are, it is a case of, “You are under arrest for X, Y or Z”.

That is how the live system works. There is no need to access any standing database such as that of DVLA; we just need the pictures of the 12,000 wanted persons. Depending on the deployment, we can choose who goes on that list. In the South Wales deployment, they were particularly concerned about terrorists who might try to cause an explosion in the arena or paedophiles who might seek to prey on the young girls going to the Beyoncé concert. The list can be varied depending on the location. I would hope, though, that any watchlist would always include people who are wanted for offences.

I will now come on to retrospective facial recognition. I repeat that we are not authorising this today by passing this clause; we are simply creating a power to make regulations in the future to do so. I just thought I would be clear about that. We are not creating this power; we are simply creating a legislative framework through which future regulations might be brought forward.

However, if they are brought forward, let me just explain how the retrospective facial recognition concept works. It is the other way round; it is where someone commits a crime and we have a picture of them. They might have been recorded on CCTV, somebody might have taken a photo of them on their mobile phone, there might be an image from a Ring doorbell, or it might be an image from a car dashcam. There is someone committing a crime and there is a photograph of them, where their face is visible, at least to some extent, but we do not know who they are. The police have a criminal—they have a photograph of them—but they do not know who they are.

What we want to do then is to search, using facial recognition, to find that person. Currently, we can search against the police custody database and we can see if that person appears on there, by matching the face. The algorithm has become a lot better in the last few months. I have seen some examples of that. I was at a meeting in Cambridge on Monday—two days ago—where some of the most recent algorithms were demonstrated to me.

There was a photograph that was about 15 years old, as in the reference photograph had been taken 15 years earlier, and the person had aged very considerably, but there was still a match. There were other examples where the person’s face was in shadow, or partially obscured, or the photograph had been taken at a funny angle. Looking at such images, we would think, “There’s no way that you’re going to match that. They look totally different,” and yet they got the match, because the algorithm now works so effectively.

I will give one or two other examples. There was a murder in a Coventry nightclub a couple of years ago and the only piece of evidence that was secured was that someone had got a photograph of the murderer in the nightclub. It was quite dark, as people can imagine, and the police did not know who it was, because they had run off; nobody knew the person’s name or address, or anything. The police ran that photograph through the database—the police national database—to get a facial recognition match. They got a name and address. The police attended the address and found the person there, with clothes that were stained with the victim’s blood. Obviously, that was the basis for making an arrest.

There was another example of someone who had committed a sexual assault while they were on a bus and the only piece of evidence—obviously, the victim gave a description, but it is quite hard to go off a description—came from the CCTV on the bus. There was video footage of the person committing the sexual offence, or sexual assault, while they were on the bus. It was run through the database and there was a match; the police found out who it was, went in and arrested them, and they are now in prison.

On a technical point, could this technology also be used to find witnesses on the bus, for example? I do not expect the Minister to know the answer to that question, but just to add a bit more enthusiasm to his love of this technology, I can see that there might be such uses for it.

That is a really good question and I do not know the answer to it. However, I can certainly find out and come back to the Committee. I know that this technology can be used to find missing persons. If someone has gone missing, the police can do a search and see if they pop up somewhere.

As for finding witnesses, of course the witness has not committed a crime, so one would have to be a little bit careful about infringing on their rights. However, I will take that question away and get an answer, and I will write to the hon. Lady and the shadow Minister, and indeed the rest of the Committee, because it is a very good question. I am afraid that, as I stand here today, I do not know the answer, but that could help to find witnesses.

The Minister is being very generous with his time. There is a degree of comfort with the point about regulation. If concerns continue to grow about the use of DVLA data in retrospective facial recognition, Parliament will have its day. And on that basis, I will not divide the Committee.

However, I want to press the Minister on something else. He is right to say that this clause is an enabling clause, but he must have an idea of what he wants it to enable. Why is it in the Bill? The Government must have something in mind for it; otherwise, there is no point in it standing part of the Bill.

I think that I have just spent the last five minutes or so—

So it is for that, then.

It could be; that is certainly a possibility that I have in mind. If those regulations were brought forward, and maybe there would be some further clarifying codes of practice and so on that will be needed as well, it would in theory be possible to bring forward regulations such that, if the person or the suspect—it might be the murderer in that Coventry nightclub, or the sex offender on that bus. Imagine if their image was not on the custody database; imagine that they had never been arrested, or convicted, or anything. The police did not have their picture; they would not have found them. It would obviously be useful, particularly for more serious offences, to be able automatically to search other facial image databases besides the custody one, in order to catch the rapist, the serious sex offender and so on.

The hon. Member for Birmingham, Yardley asked a question about her son. There is a question of proportionality here. For example, regarding this database, in these regulations one could—this is not for the Committee today, but I am just saying it—specify a threshold, and say that this would only be used for crimes over a certain level of seriousness. Obviously, rape and murder would meet that. GBH, drug supply, firearms, knives and so on, might also be specified, but a parking ticket would probably not be. In these regulations—this is one of the reasons why the wording is flexible—one could specify that threshold, and we would obviously then debate where that balance should lie. One does have to be cognisant of proportionality and privacy considerations, as indeed data protection law in any case requires us to do.

I hope I have given the Committee a bit of a flavour of how these regulation powers—

An education, Minister!

It is not often that someone says that after I have been speaking. I hope I have given a bit of a flavour of how this could help catch serious criminals who might otherwise not be caught. We would debate the specifics when the time came. There would, of course, be a code of practice and everything else to set out those guard rails. I have already indicated in response to, I think, the hon. Member for Bootle, that there is a case for a wider policy statement setting out the strategy in this area, which we can discuss and debate.

I appreciate that we have strayed a little beyond the strict confines of this clause, Ms Bardell, but the Opposition raised some reasonable questions that I really wanted to try hard to answer. I will finish by saying that I spoke to my local police commanders, Superintendent Mitch Carr and Chief Superintendent Andy Brittain, encouraging the deployment of live facial recognition in Croydon. I know 17 people got arrested. We have heard how it was done in Essex: in Southend and in Chelmsford. I encourage members of this Committee to talk to their local police and say, “Try it out in your area,” and see who gets caught. At the moment, South Wales police and the Met have the necessary vans, but they will lend them. I think South Wales police lent the vans to Essex. If anyone wants to suggest it, they could ask their chief constable to borrow the van from South Wales police or the Met, try out in their constituency, and see who gets caught.

Amendment 33 agreed to.

Amendments made: 34, in clause 21, page 18, line 19, leave out “police force listed in subsection (4)” and insert “body”.

This amendment and amendment 35 specify in more detail the relevant chief officers of bodies to which the provisions will apply.

Amendment 35, in clause 21, page 18, line 20, at end insert—

“(aa) in relation to a police force listed in subsection (4)(b) to (d), the chief constable of that police force;

(ab) in relation to a service police force listed in subsection (4)(p) to (r), the Provost Marshal of that service police force;

(ac) in relation to the tri-service serious crime unit, the Provost Marshal for serious crime;

(ad) in relation to the Service Police Complaints Commissioner, the Service Police Complaints Commissioner;”—

See the statement for amendment 34.

Clause 21, as amended, ordered to stand part of the Bill.