AI & the future of policing and security

A major investigation can run to terabytes: phones, chat messages, camera footage, cloud storage. What is new is that AI connects different kinds of investigative material. It links a name in a chat to a phone number, that number to camera footage at a petrol station, and that footage to a location on the night of the burglary. It recognises faces, vehicles and objects in hours of video, makes conversations searchable and ranks material by relevance. Where the data can be accessed quickly, a first analysis could soon be ready before the interview starts. The judgement stays human: which message is evidence, which detail clears a suspect, and what does the defence need to see?

Organised crime runs on money, and money leaves a trail. AI sorts thousands of transactions, spots patterns such as repeated small deposits and rapid onward transfers, and maps networks of accounts that keep dealing with one another. Many cryptocurrencies record transactions in a public ledger, and analysis software reconstructs long chains of transfers within it. With instant payments, money moves faster and faster. Banks already flag suspicious patterns as they happen. The next step is for signals from several banks to reveal a laundering network together while the money is still moving, within the legal limits on data sharing. A mapped network is not yet evidence: the detective establishes who knew what was going on.

An AI agent is a system that carries out a series of steps on its own to finish a task. In the control room, AI now transcribes, translates and summarises in the background. What is new is handling calls independently: some control rooms let AI deal with non-emergency calls. One control room saw the number of administrative calls reaching staff fall by about a third, and stay there. The next step is a full picture of the situation during the call. Available camera footage, sensor data and earlier reports at the same address then come together, so the call handler knows more while the caller is still talking. The call handler decides which help is sent.

Now that anyone can make a convincing deepfake with AI, a judge wants to know how a video or audio clip came about. Investigators combine detection software, which looks for traces of editing in a face or voice, with analysis of the file data. Alongside exposing fakes, a second pillar is growing: checking provenance. Some cameras and phones already add a digital seal of origin, with signed information about the recording and any edits. That helps assess where a file came from, but does not prove that what was filmed really happened that way. As more devices add such a seal, provenance analysis could play a far larger part in digital evidence in court by around 2040.

Some places are too dangerous or too large for people to explore first. Drones and robots go ahead instead. A drone with a thermal camera searches a forest at night for a missing hiker, and AI picks out a human shape in the footage. At a crime scene or after a crash, a drone produces a three-dimensional map in a short time. In some places a drone takes off first when a call comes in, giving the control room live images before a patrol car arrives; a first controlled trial showed shorter response times. For a suspicious package, a robot goes in to take a look by remote control. The officer decides how to act on what the machine sees.

Some organised crime blends into ordinary society: drug money disappearing into property, a restaurant or a car dealership, or a business owner put under pressure. Such networks are rarely visible from inside one organisation. The local authority sees a licence, the tax authority an unusual return, the bank an odd flow of money. Only together do they form a picture. AI can combine those signals and show where they overlap: the same address, the same directors, the same money flows. Joint data processing is subject to specific legal conditions. A pattern is a lead, not a suspicion: an analyst assesses the sources and what the link means before anything else happens.

AI-first means that AI and new technology do the work people do not need to do, including repetitive and predictive work. Officers and detectives put their effort where their authority, judgement and contact with people make the difference. AI carries out recurring tasks and compares large volumes of data. In one possible picture of 2040, rosters, forecasts and first analyses run continuously in the background. Ahead of a busy day, AI compares the options for deployment, leaders choose and officers do the work on the street. Whether that really frees up time depends on how the work around it is organised: AI simply bolted onto an old process often delivers less than expected.

Digital detectives and drone operators are already a normal part of the police. What is changing often has no job title yet. Guiding AI agents becomes a responsibility in its own right: a detective divides lines of inquiry across different systems, compares their results and works out why they contradict each other. Specialisms that are small today are growing as well. Synthetic media analysts establish whether a recording is genuine or made by AI. Virtual world investigators follow crime in environments where offenders operate as avatars. Lawyers assess whether the output of an AI system will stand up in court. Not every task becomes a new profession, but a career in investigation increasingly starts in the data.

In an investigation team that works with AI, it is mainly the human work that changes. AI helps search large volumes of messages, images and statements, builds timelines and points out contradictions between witnesses. The detective becomes more of a director than a searcher, sharpening the lines of inquiry, choosing which scenarios to pursue and questioning what the system produces, including anything that might clear a suspect. Talking to a witness, interviewing a suspect and weighing a statement remain human work, and grow in importance. That calls for different skills: reading critically what a system delivers, and knowing when a connection is too neat to be true. With more connections in view, an experienced detective can ask sharper follow-up questions and open new lines of inquiry.

Police leaders face close public scrutiny, which is exactly why their example counts. A shared picture of the future gives direction: what will police work look like in ten years, and what does that mean for the officer starting today? Leaders take the initiative here. They start using AI and new technology themselves, show what it delivers and explain why. That is how they inspire officers and detectives to help build it. Their example makes the change tangible: people see where the leadership wants to go and how they can contribute themselves.

A future academy is where an organisation trains its own employees and leaders of the future. The police do not wait for education to catch up: the teaching material comes from their own work. A closed case, an operation that went differently than planned or an interview that stalled becomes a case study in which officers, detectives and new recruits practise with AI and new technology. They learn what the technology can do, where it falls short and how to use it in their own work. Leaders decide together what they leave to AI. That way, the organisation's knowledge grows with the technology.

An ecosystem means that police teams at home and abroad, start-ups, scientists and partners form a single whole, with shared analyses and jointly developed technology. Problems ignore the borders between organisations and countries, and so do solutions. Partners can test each other's applications and adapt them to their own data and ways of working. With federated learning, partners already train AI models together while the data stays with each organisation. As more forces and countries work together like this, an innovation reaches the street sooner.