Predictive policing

WHAT IS PREDICTIVE POLICING AND HOW IS IT BEING USED?

Predictive policing is a type of artificial intelligence (AI) tool that is used by the police forces across the UK to predict where and when a crime will be committed, who will commit it, and even who the victims might be.

For years, these tools have escalated in scale throughout British police forces and are already influencing decisions about people’s freedom, safety, and legal outcomes with little governance in place.

Liberty Investigates and WIRED recently revealed that, in Bristol, over 90% of the population have had their data fed into predictive models that assigned risks scores to adults and children and create a ‘picture of threat, harm, and risk in the region’.

These systems are built on highly sensitive information – including mental health records, housing status, teenage pregnancies, and if a child is receiving free school meals.

Liberty Investigates went on to reveal how two of these predictive models were used by Bristol City Council to assess risk of child sexual and criminal exploitation. They were quietly shut down after staff described them as “not fit for operational use”.

Avon and Somerset Police have spearheaded the use of this predictive technology for the last 10 years. They created dozens of predictive systems, including models to anticipate who would commit burglary, become a victim of domestic abuse, or go missing.

WHY SHOULD WE BE CONCERNED?

Liberty has three core concerns:

1. that police are using personal and public service data without people’s knowledge,
2. that untested and potentially unsafe AI systems are being used by police without communities’ understanding or consent,
3. and that people have no means of knowing when an AI system has made a decision about them – and therefore no way to challenge it.

The public is given very little information about what AI systems are used and how predictive algorithms reach their decisions – and even the police don’t understand how they do it. This lack of transparency and understanding means these programs can’t be properly scrutinised and life-changing decisions are near-impossible to challenge.

At the start of 2026, the Government created Police.AI, a new body which will have the expertise to scrutinise police AI systems and help roll out tools to 43 police forces in England and Wales. However, to be effective, it needs to have power. Most importantly, it needs an overarching legislative framework that can adapt to the pace and risks of emerging technology.

WHAT ARE WE CALLING FOR?

Governance and oversight for these systems is long overdue.

That’s why we’re calling for robust safeguards and oversight on when police can use these tools, how our data is used and protected, and who is accountable for the decisions these systems make.

We’re specifically calling for:

Tell the public: Police must be fully transparent about their use of AI

Clean data, clear limits: Binding rules on what data police can feed into AI systems

Test before deployment and keep testing: Independent testing must be mandatory before any AI system is used on communities, with continuous monitoring and a ban on systems found to be unsafe or discriminatory.

A human must always be accountable: Every AI-assisted decision must have a named, accountable human behind it and clear consequences for misuse.

Train the officers using it: Every officer using AI must understand how it works, its limitations, and their legal responsibilities when acting on its outputs.

If you can’t audit it, you can’t use it: Every AI system used by police must produce a clear, traceable record of how decisions were reached.

The right to know and to challenge: Anyone affected by an AI-assisted decision must be told and given enough information to understand it and challenge it.

A regulator with real teeth: The Government must empower Police.AI to be an independent, properly resourced regulator to enforce every one of these standards.

If you agree, join our campaign today.