The Hidden Algorithms Behind Modern Policing

In this episode we discuss something that sounds like science fiction, but is already part of everyday policing in the UK; predictive policing. These are tools that use data and algorithms to help the police forecast crime, where it might happen and sometimes who might be involved. The idea is to use resources efficiently and cut crime. And there are reports that councils are also starting to use these systems (The Think Family Database.)

But a recent report by Amnesty International, “Automated Racism” argues that predictive policing tools aren’t neutral; they may be reinforcing and scaling existing inequalities. The report argues that the data they use is biased, particularly against black and racialised communities in deprived areas. 

In this conversation, we unpack what these tools actually are, how they’re being used, whether they work, and what the risks are, especially when combined with other technologies like facial recognition.

Our guest is Ilyas Nagdee who is a human rights campaigner and the Racial Justice Director at Amnesty International UK. He has also written for The Guardian and is the co-author of a book entitled, Race to the Bottom: Reclaiming Antiracism, a critical study of anti-racist politics in the UK.

Useful Links

Phaselaw

Amnesty International

Automated Racism Report

The Think Family Database

This podcast is sponsored by Phaselaw - a purpose-built solution for document disclosures, like subject access requests and FOI requests. Instead of redacting PDFs one by one, or forcing litigation software to do a job it wasn't designed for, with Phaselaw you get collection, review, and redaction in one workflow. Teams across the World are using it to cut response times from weeks to days.

For Guardians of Data listeners, Phaselaw is offering a two-month free trial; run it on live requests, see what it does to your backlog, decide from there. No card, no commitment.

Transcript

Ibrahim Hasan 0:06

Welcome to Guardians of Data, the show where we explore the world of information law and information governance, from privacy and AI to cybersecurity and freedom of information. I'm Ibrahim Hassan. Today we're talking about something that sounds like science fiction, but is already part of everyday policing in the UK. Predictive policing. This involves the police using data and algorithms to try and forecast crime, where it might happen, and sometimes who might be involved. The idea is to use resources efficiently and cut crime. But a recent report by Amnesty International entitled Automated Racism argues that these predictive policing tools aren't neutral. They may be reinforcing and scaling existing inequalities. The report argues that the data they use is biased, particularly against black and racialized communities in deprived areas. This is leading to the policing of these communities based on who they are, their backgrounds, where they live, and who they associate with. So in this conversation, we're going to unpack predictive policing, whether it works and what the risks are, especially when combined with other technologies like facial recognition. I'm joined by Ilias Neddi, who is a human rights campaigner and the racial justice director at Amnesty International UK. Ilias works on issues including systematic racism, policing, protest rights, and civil liberties. He has also written for The Guardian and is the co-author of a book entitled Race to the Bottom: Reclaiming Anti-Racism, a critical study of anti-racist politics in the UK. Let's jump in. Ilias, thank you very much for joining us. I'm really excited to hear what you have to say.

SPEAKER_00 1:59

Thank you very much for having me.

Ibrahim Hasan 2:01

Let's start with the basics, because I think a lot of people don't fully understand what predictive policing is. It's a bit of a vague concept. What are we actually talking about in practice in UK policing today?

SPEAKER_00 2:15

Yeah, uh I think you're absolutely right. Many people might look at a term like predictive policing, and they might actually think it's a system that's much more sophisticated than it sounds, right? So the first thing to note is there's no agreed formal official definition of predictive policing in the UK. Uh, there's not even the definition that's sort of widely agreed upon. When you look at the term itself, it asserts that there's a predictive function that's been carried out, and our research sort of disputes that. But as Amnesty International, we describe predictive policing systems as computer programs that use data, algorithmic models to assess the risk that a crime will be committed. So there's two main types of predictive policing systems that are used by forces in the UK today. So one is person-based predictive policing, and that's assessing the or calculating the risk score that someone's going to be a victim or a perpetrator of a crime. And the second is place-based predictive policing, and that's assessing that a location will be at the site of a future crime. So those are the two sort of systems that exist.

Ibrahim Hasan 3:24

Can you walk me through a concrete example, Ilias? What kind of data goes into one of these systems and what are the outputs?

SPEAKER_00 3:32

So one of the examples that we talk about in the report is a place space, a predictive policing system, which is the West Midlands Police knife crime prediction tool. That's been in use for a number of years. The force says that the aim of that system is to predict where and how often ice crime might take place. So it uses historic data, uh, data and trends from the last 20 years, from existing police data, and it divides the location into sort of one kilometer square grids on the computer system. And then they identify places for police patrols. These areas are smaller than sort of a council award. You're looking at streets or a few streets. And what we found, and what's been clear, is actually where police are carrying out enforcement actions are in places around like Soho Road and Rosels in Birmingham, following predictions from the system. And those are areas with higher numbers of ethnic minority communities, racialized communities compared to the rest of Birmingham. And it's also one of the most deprived areas in the city. And when you look at an outcome of a system like that, what West Midlands police have themselves acknowledged is it has predicted knife crime correctly, less than one in five times. So it's been incorrect, sort of more than four in five times. What's happened in the West Midlands police area in the time that this system was in use is there's been disproportionate numbers of stock and search of black and Asian people compared to their white counterparts in those areas. There's been particularly higher levels of suspicionless stock and search, so Section 60 stock and searches in those areas. And there's also been disproportionate use of force against people in the area that West Midlands Police covers. And the police themselves have said that they can't satisfactorily account for that disproportionate use of force. But what we would say is as a result of that system being in place, police officers are going into areas with the mindset this is where crime takes place. And of course, wherever crime is thought, crime will be found is just what these systems are doing, putting these communities under the magnifying glass and placing them more at risk of not only disproportionate levels of police intrusion through stop and search, but also disproportionate use of force.

Ibrahim Hasan 6:03

That's interesting. Just to summarize, they're using historic data about crime to target particular areas and to predict where crime may happen. The police might say they're just making better use of resources. And if it leads to better policing and more law enforcement, surely that's a good thing.

SPEAKER_00 6:26

Yeah, so this is what we look at in the report through a concept known as a feedback loop. So what that means is police and you know, many police forces in the UK have come forward and said that there's institutional racism and discrimination within police forces. We know, especially when you look at documents like the Casey Review, the Macpherson Report, and others, that police forces across the country have historically and longwithstandardly disproportionately policed particular communities. And is that data is as a result of where police have used data that reflects the structural and institutional discrimination present in policing and the criminal legal system. So you know, the way that police say communities will have been disproportionately policed in the 90s, the hundreds, that information is being used to make predictions or create profiles. That data includes things like police intelligence, stock and search, arrest data. And of course, we know historically where areas where there have been high populations of black and racialized people have been disproportionately policed. So all that's happening is the bias that exists in that data is leading to predictions that crime will occur in those same areas, or individuals from those backgrounds are likely to commit crime. So police then return to those areas through the guise of, well, that's where the system's telling us we need to go. And because they're going to those areas compared to others, that's where they're finding crime. So that goes back into the system, which is sending them back into that area. So it's a repeating cycle and feedback loop that's occurring.

Ibrahim Hasan 8:03

Some of those messages are going to really resonate with our listeners, a number of whom are data protection and privacy practitioners. The idea that historical data is being used to feed into a system to make decisions about the future. And of course, if the data itself is not accurate, hasn't been fully checked, then it's going to perpetuate existing biases and compound and almost industrialize these biases. Is there any evidence that AI is being used in this system as well?

SPEAKER_00 8:36

So, in terms of AI, specifically, it wasn't like a major feature of the research report. So I don't know if I could speak to that fully. We know that particularly government proposals over recent years have really pushed with the idea of utilising AI increasingly within not only policing but also other areas of the criminal justice system and more widely across the public sector as well. There is like some stuff which crosses over, but it wasn't like a major feature of the research report.

Ibrahim Hasan 9:07

But I suppose we can predict that AI will soon feature in these systems. And so that will perpetuate even more bias and even more disproportionate impact on certain communities.

SPEAKER_00 9:21

Yeah, absolutely. I think we come with reasonable confidence when looking at the systems that are in place already and the outcomes they're having, say that it is very likely that the use of AI in these systems is going to continue this sort of compounding cycle of discrimination that's felt by communities and individuals within those. And the other thing that not only AI introduces, but what is currently the situation as well, is just the lack of regulation. In so many areas, there is an open acceptance that legislation just has not managed to keep up with advances, whether that's within social media, in AI, and in other things. And I think predictive policing is a prime example of an absolute absence of regulation.

Ibrahim Hasan 10:08

And just to clarify, Elias, are these systems mainly being used to predict places, or is there evidence to suggest that they're being used to target individuals as well?

SPEAKER_00 10:19

Yeah, so I mentioned that there's sort of two types of systems earlier. So there's a place-based and the people-based. So one of the examples of a people-based predictive policing system is one that's used by Avon on Somerset Police called ClickSense, and that through the ClickSense platform, including the system, the Offender Management App, which is designed to monitor risk levels of offenders. And this operates similar to your listeners might get familiar with one of the most famous examples is the Metropolitan Police's Gangs Matrix. And the Offender Management App in use in Avon and Somerset Police Force area uses historical data from police recorded crimes to profile people linked as an offender to crime. And what it does is it gives you a score between zero and a hundred to create an individual sort of risk score and harm score. But the force defines an offender or being linked to crime as including things like, for example, just being a suspect or even a possible suspect, or being issued a fixed penalty, having a community resolution or restorative justice outcome, being given a youth or adult warning or caution, or even if there's insufficient evidence, all of those things contribute to your risk score. And what we know in terms of the way that that particular system works is if your risk score is over 70, it's classed as high. If it's 40 or more is medium, and if it's below that, it's four. I think quite often a misconception around these systems is, you know, we're talking about targeted. This is targeted at specific people. Even on Somerset Police, when we were carrying out this research, we found like in their documentation that the numbers that were present said around 300,000 people were on its offender management app. As many as 170,000 had been given a risk score in the last six years. That's not a small number of people. There's been no formal evaluation report on any of its click apps. Even on Somerset Police, they say that we share this information under outputs with other agencies, including the city council and stuff like that. And just as an example of how this impacts people, we spoke to an individual who had been profiled by this system and given a risk score. He told us he reckons he's been stopped and searched over 50 times. It was only because he was stopped by police so many times that he submitted a subject access request. And he found that he'd been profiled through the system, but they wouldn't call him anymore. They wouldn't tell him what his risk score is. He found out that the police probably had incorrect data about him on his criminal record, which is probably why he was risk scored in the first place. And he asked them to correct that. And they said they weren't able to. And it took a legal challenge for him to be able to get that correction to his criminal record. And when people think about stop and search, many people may never have had to experience a stop and search. Some people may have experienced one stop and search. But imagine trying to live your everyday life and possibly many of your interactions with police, because you've been profiled on a system you don't know about, you don't know what your score is, you don't know how to challenge that, what information underpins that, how to ever get that changed. You're coming out of Tesco, you're walking to the bus stop, you're trying to do anything, living your day-to-day life. And all you're facing in that is a level of police intrusion that is disproportionate and should never be the case.

Ibrahim Hasan 13:46

Thank you for that, Ilias. You've very clearly explained that the whole argument sometimes people have when we raise awareness of these kinds of issues is well, nothing to hide, nothing to lose. If you're not a criminal, you don't have to worry about. What you've clearly explained is that we all should be worried because the people, these systems are processing information about making decisions, is more than just your person who is of interest to the police. It's it's everybody or potentially everyone.

SPEAKER_00 14:20

Absolutely. These technologies and these systems aren't just used in big cities. In our research, we conclude that almost three-quarters at the time when we published of UK police forces are using data-driven systems such as these ones. And the use of these approaches is influencing decisions in policing and the criminal legal system, and also impacting people's access to essential services. We're not talking about small parts of the country, we're talking about the overwhelming majority of the country where police have either historically used or are currently using predictive policing systems. And I think one of the points that you made there is around people thinking about the idea that if they have nothing to hide, that they can be transparent about the way that they live their lives. I would flip that argument a little bit and I would say that there's a significant lack of transparency around police use of predictive policing systems in the UK. Most people don't even know about their use in policing. And most people think of systems like this when they're thinking of like sci-fi movies or minority report or something. But they don't think about the fact that systems like this are used in policing and the impact that has on where they live, how they're affected or targeted. People don't know if if they are ever stopped and searched by police, it's as a result of a predictive profiling or risk assessment system. They don't know if they ever find out that they've got a score, how they would challenge it. Quite often what we found is when people do try to seek information, they're met with legal refusals, exemptions, rebuttals, and things like that. So there's a significant lack of transparency, which is particularly worrying given the lack of regulation.

Ibrahim Hasan 16:05

I agree. The lack of transparency is of concern. I was reading recently that the pressure group State Watch used the Freedom of Information Act to discover that the UK government is developing a murder prediction program. Minister of Justice hopes that the project will boost public safety, but of course, campaigners say that it's chilling and dystopian.

SPEAKER_00 16:29

Absolutely.

Ibrahim Hasan 16:31

You also mentioned, Ilias, the gangs matrix. I'm glad you mentioned that, because uh the information commissioner a few years ago served an enforcement notice on the Metropolitan Police saying that the system wasn't complying with existing data protection laws. So it's interesting what you've said with regards to the legislation. Do you think that the legislation needs to be updated and we need specific legislation around these systems?

SPEAKER_00 16:58

Yeah, just on the subject of the gangs matrix, as you rightly outlined after that finding from the ICO, the Metropolitan Police on February 2024 said that they were going to discontinue the use of the database. But what we're particularly worried about is what's come in, which is the violence harm assessment, which the Metropolitan Police says is an intelligence tool to prioritize and drive operational activity to focus direct resources. When you look at the outcomes of the violence harms assessment, I think the last data I've got in front of me is in August 2024, there were just over 1,300 people on the violence harm assessment. Of those, two-thirds were black. So what we can clearly conclude from that is when that's so disproportionate to the population and the ethnic breakdown of London, all that is is the gangst matrix alive by another name. So we are particularly worried about is as a result of the lack of regulation in this area as well. Even where forces have been found previously to be misusing these systems or these systems leading uh to discrimination, both in the way that they are built and the operational decisions that they lead to, these systems are just continuing with that under another name.

Ibrahim Hasan 18:17

The legislation that we have at the moment, we agree that there is no single piece of legislation. We've got the UK GDPR, we've got part three of the Data Protection Act, which applies to law enforcement agencies, and then we've got the human rights framework. Is that enough, or do you think we need further legislation to regulate this area?

SPEAKER_00 18:37

Yeah, absolutely. I mean, so I in terms of regulation, it's been quite interesting. I think this goes back to that point we were saying around transparency as well, right? When we were doing this research and we wrote to every police force in the country, there was quite a lot of resistance to those that wanted to provide information. Some just didn't respond to our FOI request, others cited exemptions. Even the Metropolitan Police Service, in their response to us, has sort of said we recognized your description of forces giving pushy responses to your queries around these systems. When we wrote to the Minister for State for Policing, Fire and Crime Prevention, they sort of told us that in terms of regulation, all chief constables had signed up to the AI Covenant for Policing. But when we look at that covenant and what it does, that's not a satisfactory safeguard in any way. When we were carrying out this research, our research project took almost two years. Police forces didn't reply, replied only when chase or prompted, uh, sometimes they didn't respond at all, cited blanket exemptions, sometimes provided contradictory information saying they didn't use predictive systems when there was other official sources and information that said that they did, failed to send documents they had agreed to disclose. The Metropolitan Police told us that their systems have independent oversight through the London Police and Ethics Panel for any contentious use of data, and they wouldn't use any predictive data system without being open and transparent about it. But there's no publicly available assessments by this ethics panel of either the Metropolitan Police Service use of TM risk terrain modeling or the violence harm assessment system I was talking about earlier. There's no mention of the policing ethics panel in the data protection impact assessments or equality impact assessments completed by the force in relation to either system. Again, when we wrote to the minister, one of the things that was cited to us is the algorithmic transparency standard, which was created by the Central Digital and Data Office and the Center for Data Ethics and Innovation as part of the national data strategy. And the aim of that is so public sector organizations can provide clear information about algorithms they are using to support decisions and where they may have an effect on the public. These systems would demonstratively fall into that category. And they said they should be made public when it's being piloted or deployed. In the algorithmic transparency record, there's only records of two systems in use by police forces when we were carrying out this research. That's nowhere near the number of systems that exist.

Ibrahim Hasan 21:19

So you're saying there's a disconnect between what the police and law enforcement agencies are saying in terms of transparency, openness, accountability, dealing with people's rights and what they're actually doing. That's interesting. If Amnesty were invited to be involved in drafting new legislation to govern this area of predictive policing, because with AI, as we've said, it's only going to increase and the power of AI is going to be harnessed to do more of this kind of stuff. What would be your guiding principles in drafting legislation to govern this area?

SPEAKER_00 21:56

I guess like the first step is all of us agree. States have a responsibility to keep us safe, to keep us, our families, our community, and all of us in this country safe, right? And there are actions and things that police can do in order to ensure that it that intrudes on our human rights. And the entire thing around there is looking at how those responsibilities are balanced and how those restrictions to our rights, our rights to privacy and family life, which are subject to certain restrictions, are provided by law, are necessary and proportionate, and are for the purpose of protecting specified public interests like national security. That's the test set out in international human rights, the standards. If states can demonstrate that those restrictions to our rights that I've spoken about and that we document in our report are for the purposes and meet this test, then those are legitimate restrictions. But what we have found, and our research concluded, that the police use of these systems necessitates widespread monitoring, collection, storage, analysis, or other use of personal data, including specifically sensitive data. It does so without reasonable suspicion of criminal wrongdoing. And we believe there's evidence in the UK about the disproportionate targeting of racialized communities, people from deprived backgrounds. And what that does is that meets the standard to be described as indiscriminate mass surveillance, which can never be proportionate interference with several of our rights, whether that's privacy, expression, association, assembly. So as Amnesty International, our position is all indiscriminate and mass surveillance fails to meet the test of necessity and proportionality and therefore violates international human rights law.

Ibrahim Hasan 23:50

And those key principles of necessity and proportionality are very much baked into existing data protection legislation. So even in the absence of specific legislation, if would you agree authorities, law enforcement agencies just complied with existing data protection law, that would be a good start in achieving that trust, that transparency that this area requires?

SPEAKER_00 24:18

Yeah, so I mean our fundamental conclusion as a result of it meeting the standard of indiscriminate mass surveillance is that predictive policing systems, as has happened in many places across the world, um, should be prohibited.

Ibrahim Hasan 24:31

I see. You mentioned that these systems are being used across the world. Have you seen any country in the world that is doing it right the way amnesty would expect?

SPEAKER_00 24:42

Most people think of predictive policing, and quite often what's said by even some of the firms that are selling these technologies to UK police forces is predictive policing is something that happens there. And that means specifically the United States, which, you know, there's been much criticism in the US for these particularly geographic predictive systems of repeatedly targeting poor areas and areas with high concentrations of uh particularly the states of African American people and other racialized communities. A landmark study in 2016 in California on Predpol, which is a major geographic predictive policing software company, said that the algorithm would result in the targeting of black people for alleged drug crime at twice the rate of white people, despite all the data showing that people of all races use drugs at similar rates. Another study on the same system on Predpol, by the founder of Predpol itself, found that Latino people in Indianapolis, Indiana, would have been subjected to 200 to 400% the amount of police patrols as white populations had it been deployed there. What we've looked at, and in 2021, a study that looked at 7.4 million predictions from Predpol's own crime prediction data, which had been left on cloud storage accessible by an open link, found that neighborhoods that the software targeted for increased patrols were more likely to be home to black people, Latinos, families, and particularly areas where there was higher qualification for freer, reduced lunch programs. So again, areas with high levels of deprivation. In the States, what's actually happened is a number of US police forces have stopped using predictive policing systems altogether. Palo Alto police department said they didn't find it effective and didn't get any value from it. It didn't help the department solve crime. Other police forces also stopped using it. Police forces in certain in some cities, Oakland, New Orleans, Pittsburgh, had to be forced to stop using them after local city councils banned the practice. So rather than looking at tweaks and changes that could be made, we're hopeful in moving to a space where many of these systems are prohibited in their use because of not only findings around ineffectiveness, but particularly the concerns that continue to be raised that these systems are just reproducing and compounding discrimination for communities.

Ibrahim Hasan 27:02

It's interesting you mentioned the US and the fact that some police forces have voluntarily stopped using these systems because they've realized that there are serious accuracy and data quality issues. Do you think there's a greater need to raise awareness of those issues amongst the police here in the UK to get them to fully understand the difficulties and to hopefully voluntarily stop using these systems if they are going to have an impact not just on service delivery but also the trust of the people? Because in the end, in the UK, we rely on policing by consent, don't we? If we lose the consent of the public and the trust, then it's going to make law enforcement more difficult. Do you think that message has got across or needs to get across more strongly to the police?

SPEAKER_00 27:51

So I don't think that message has got across. I think what I'm particularly worried about is the way that these technologies are often sold to services like the police and to other elements of the public sector is just as all about increasing efficiency. We're operating at a time where state infrastructure is operating with increasingly depleted resources on services that serve the public, the NHS, schools, local authorities, police forces, et cetera. And what's quite often said is these technological developments will support with depleted resources. When you look at the amount of money that's wasted on these systems, and it's not small amounts, millions of pounds have been given not only by central government to police forces, but also police forces themselves are spending, in some cases, hundreds of thousands of pounds per year to operate these systems, even when they're not having any demonstratable impact. A good example is Kemp Police. They were one of the early adopters of Predpol, that system from the US in the UK. They used it for five years between 2013 and 2018. They themselves said we're the first force in England to introduce predictive policing systems. Internal evaluation in 2014 said that it was costing the force about £100,000 a year. And the same evaluation said the results did not show an overall drop in crime for the force. At the end, in 2018, Kemp Police stopped using the system as they said it was challenging to prove whether it had even helped the police reduce crime. So police forces are being sold dreams in some cases or are chasing this use of data in the hope that it will lead to some impact that demonstratively does not seem to be occurring.

Ibrahim Hasan 29:32

That's fascinating, Ilias. Thank you for that. Just finally, I want to widen the lens a little bit because predictive policing isn't the only technology being used. We're seeing increasing deployment of live facial recognition. Recently, Big Brother Watch lost a challenge to the use of facial recognition by Metropolitan Police. When you combine prediction with real-time identification in public places, are we moving towards a form of continuous surveillance?

SPEAKER_00 30:01

So Amnesty for many years has done research and we're very public about our opposition to life facial recognition, not only because of the lack of regulation in the UK and the fact that it violates several rights, but also the angle where it's it's demonstratively, and I think even previously the Home Office themselves have concluded that the outcomes from life facial recognition pilots and the way that those systems operate discriminate against particular communities and have less accuracy, particularly with racialized communities and especially black people. So I think you're right. We sadly are moving to a time where actually many members of the public may not necessarily be completely aware with how not only the surveillance is occurring, but also we're seeing in real time the capture of our data not only by police forces and state authorities, but also increasingly and worryingly there are concerns about the role of private actors within our data collection. And I think that's really clear at the moment with the campaign for Rob Palenter and the NHS.

Ibrahim Hasan 31:08

Fascinating. Just finally, Elias, our listeners are no doubt going to want to look at this area in much more detail and read about your work. Where can they do that?

SPEAKER_00 31:19

Yeah. If your listeners want to find out more about the systems, what they do, how they operate, their outcomes, the assessment of that against human rights standards. If they just Google amnesty automated racism, that'll take you to the report that we found. But there's actions that they can take, including signing out petition. There's also recommendations that they can look at that we have made to government around how, well, firstly, our foundational and core argument is that these systems should be prohibited. And then we also have minimum sort of recommendations where we're calling on incremental steps towards prohibition.

Ibrahim Hasan 31:58

Fantastic. Ilias, it's been a fascinating conversation. Years ago, we used to say that we are sleepwalking into a surveillance society. I think now we are waking up to a society where algorithmic decision making is normal in policing. The issue very much for me is that we're doing it without transparency and without a public debate. So this kind of conversation with experts like yourself will hopefully create debate and awareness. So I'm really grateful for your time. And hopefully, we can invite you again.

SPEAKER_00 32:29

Thank you very much for having me. And thank you to you for the work that you do in helping convey these really important aspects. And sometimes, sadly, within the public debate, undercommunicated like the really important aspects of how our data and information is used and the regulations surrounding that, I think, is stuff that people really do need to know more about. And I think the work that you're doing on this podcast and and through other means is is crucial in informing that public debate.

Ibrahim Hasan 32:58

That's all for today's episode of Guardians of Data. Once again, a huge thank you to Ilias Nagdi for sharing his insights. We've learned that predictive policing isn't some distant or experimental idea. It's already embedded in policing in the UK and indeed across the world. It promises efficiency and objectivity, but as we've heard, the systems are built on historical policing data that can reflect and reinforce existing inequalities. So tools designed to make policing smarter may also be making it less fair, especially for communities that are already overpoliced. And when the police start layering these systems with other technologies, such as live facial recognition, the risks multiply and we have the possibility of continuous data-driven surveillance in public places. It's not surprising then that Amnesty International are calling for the police to stop using predictive policing. Some might say, though, that the horse has already bolted. In a cost-conscious, data-driven public sector, predictive policing is here to stay. So the challenge for information governance professionals, especially for those working for law enforcement agencies, is to ensure that the systems underpinning predictive policing comply with the law, particularly part three of the Data Protection Act 2018 and the UK GDPR. The important principles that Ilias discussed of fairness, transparency, and data quality, are already built into this legislation. IG professionals have an important role in ensuring that these principles are baked into predictive policing systems at the design stage. This requires them to be involved before such systems are deployed so they can ask the right questions and conduct the all-important data protection impact assessment. This will not only achieve legal compliance, but will also help to ensure that policing techniques maintain the trust of all communities. If you found today's discussion useful, please subscribe, share this podcast with colleagues, and leave us a review. And remember, whether you're a seasoned professional or just starting out in information governance, there's always more to learn, and we'll be here to help you stay ahead of the curve. Thank you for listening and join us next time on Guardians of Data.