Learning from the world: How global innovations can shape UK policing
Policing around the world is evolving quickly. Today’s officers face a landscape shaped by complex challenges, many driven by the volume of digital evidence now central to investigations. From cybercrime to digitally coordinated criminal networks, the digital era demands more than just better tools, it requires a shift in how policing itself is approached.
The UK has much to learn by looking outward. Around the world, forward-thinking initiatives are redefining what effective policing looks like. Two exciting projects, Australia’s Söze platform and the Netherlands’ National Police Lab AI, demonstrate how carefully guided technology can set new global standards.
At the heart of the Netherlands approach to modern policing is the National Police Lab AI (NPAI), a groundbreaking partnership between the Dutch National Police, Utrecht University, and the Delft University of Technology.
The Lab’s origins lie in a simple but urgent reality: the Dutch police lacked internal expertise in Artificial Intelligence (AI). Recognising this gap, the NPAI was created not just as a hub for developing cutting-edge tools but to promote the possibilities of AI within the police force, including among senior leadership. The goal has been to make policing safer, smarter, and fairer by integrating scientific knowledge into how technology is evaluated and deployed.1
The Lab is led by AI expert and researcher Dr Bas Testerink, whose background in computer science and links to academia has been pivotal. His current role at the NPAI spans research, strategic engagement, and building a bridge between operational policing and the universities’ technical capabilities.
Unlike typical R&D setups, the NPAI is embedded within the Dutch police system but maintains strong academic grounding. This dual identity allows the Lab to combine the technical expertise of academia with the practical demands of day-to-day policing. For example, some PhD students from Utrecht University are situated in secure, purpose-built police facilities, giving them direct exposure to real-world challenges while maintaining necessary safeguards around sensitive data.
“Some of the PHD students from the Lab are housed in one dedicated police room. It’s like a separate security system” Dr Bas Testerink said.
The NPAI is multidisciplinary by design. It brings together data scientists, legal experts, AI ethicists, communication experts, and operational police staff. Developing new AI models is just one element of a broader process which includes legal frameworks, user experience, and communications.
The Lab’s work is supported by an Ethics Group and an Implementation Group. Their role is to make sure that new technologies are not only functional but deployable, understandable, and aligned with societal values. Ethical questions about whether AI should be used in new scenarios are embedded throughout and not bolted on at the end. This means that solutions are crafted with real-world application and public trust in mind.
The Lab operates through both centralised and decentralised models. Centralised projects are major systems and operate through structured pipelines and dedicated software teams. These require long lead times and formal planning cycles, so they are robust, enterprise-ready, and fully integrated with legacy systems.
Decentralised development on the other hand has emerged more organically over recent years. For example, investigators and analysts embedded in local teams often hire their own data scientists to develop bespoke tools quickly, responding to specific police operational needs. These data scientists work directly with frontline officers rather than relying on the centralised IT pipeline, allowing for more agile and practical tool creation.
The Lab sits between these two worlds, formally part of the centralised structure, but practically connected to the frontline. This hybrid model means it can provide deep research, prototype tools, and ensure they are usable and adaptable in the field. For instance, when a digital crime specialist needs a tool but lacks the time or resources to conduct extensive research, the Lab can step in to explore new methods, evaluate the latest research, and develop viable solutions.
Before a new tool is even prototyped, ideas are evaluated to assess strategic fit, legal viability, data availability and user need. If viable, a project moves through a structured “stage-gate” innovation process, with multiple phases of user testing and refinement before reaching full deployment and permanent maintenance. This method ensures it remains practical and compliant from day one.
One of the Lab’s most high-profile successes is MonoCam - an AI-enhanced camera system that detects drivers illegally using mobile phones. It’s a blend of computer vision and human oversight. The system flags images, which are then confirmed or dismissed by a human operator (Figure 1).
Figure 1. Screenshot of the MONOcam software interface as shown on the laptop. The top left shows a livestream of incoming video, the top right shows the photographed car, with evidence of movement. The bottom right shows photographs of all passing vehicles. Potential hits are added to the list on the bottom left for further manual inspection by the patrol officer. Non-hits are deleted.
Donatz-Fest, I. C. (2024). Values? Camera? Action! An ethnography of an AI camera system used by the Netherlands Police. Policing and Society, 35(1), 50–67. https://doi.org/10.1080/10439463.2024.2370939. Deed - Attribution 4.0 International - Creative Commons
However, the Lab’s value lay not just in developing the model, but in studying how it worked in practice. A PhD student observed that user boredom led to problematic behaviour, prompting changes in the system’s design to limit exposure times and reduce the potential for data leaks.
This kind of observation-led refinement grounded in behavioural insight and ethical awareness is what distinguishes NPAI’s approach. It’s not just about what the technology can do, but what it should do, and how it fits into the realities of human use.
But not every problem needs an AI solution. One of the Lab’s most important contributions has been challenging the automatic assumption that AI is always the answer. This surfaced a raft of unmet operational needs, many of which are solvable with simpler, existing tools. This has resulted in better informed deployment, improved resource use, and a more thoughtful, evidence-led approach to innovation.
As Dr Testerink states “You don’t always need the latest AI to make a difference, sometimes the right solution already exists, but people just aren’t aware of it”.
There are key lessons here for the UK should it consider a similar model.
Modern policing faces a daunting challenge with the sheer volume of digital evidence pouring in faster than officers can process. In response, Western Australia Police, working alongside the global digital engineering company, Akkodis developed Söze, a powerful tool designed to simplify complex digital investigations. Söze quickly gathers large amounts of digital data into one clear, easy-to-use system. It can instantly sift through evidence from different forensic sources and handles 137 languages, making it effective across borders in international cases.2
Unlike many AI tools, Söze does not rely on Generative AI, avoiding the risks of fabricated or false data. Instead, it acts as augmented intelligence, boosting human analysis without replacing it.
Söze empowers investigators with powerful analytical capabilities. The platform enables investigators to simultaneously access, tag, and analyse extensive volumes of complex data including video footage, call records, financial transactions, and social media, rapidly uncovering insights that traditional methods cannot.
By automating complex data correlations across multiple sources simultaneously, Söze helps investigators to pinpoint patterns and connections that were impossible to detect manually or required substantial policing hours to comb through the data. The system has been instrumental across diverse crime types including terrorism, organised crime, homicide, human trafficking, cybercrime, and child exploitation, fundamentally transforming the speed, precision, and success rate of investigations.
Since summer 2023, Avon and Somerset Police has been piloting Soze , assessing how it compares to traditional investigative methods. Investigators report that Söze uncovers key evidence within minutes, some of which previously would have been buried under mountains of data. For instance, during an initial six-week trial in 2023, Soze was able to process 261GB of data across 27 cases in the equivalent of just over 1 day -which would have taken 81 years using traditional methods.
Police officers quickly adapted to the Söze platform during the trial phase, mastering around 80% of its capabilities within just 90 minutes of training. The simplicity of the system for users enabled Avon and Somerset to quickly put the tool to work and deliver tangible results in ongoing cases.
“It gave us confidence from the beginning not just because of the time it could save us, but because it supported the level of investigative quality we rely on”.
Edward Yaxley, Detective Superintendent at Avon & Somerset Police.
Proof of value and financial implications: The 12-month proof of value trial builds on an initial six-week trial. Avon and Somerset Police are assessing the overall impact on policing outcomes by evaluating investigative time savings and evidence quality in supporting briefs for the Crown Prosecution Service. Söze’s ability to meet the specific requirements of UK policing and data formats will also be assessed, along with how the tool can be accessible to officers in their daily work environments, and the effects on staff satisfaction and retention rates.
Once the proof of value trial ends, a final decision will be made on whether to proceed with broader integration of the tool. Given the substantial costs associated with IT projects and data management, the decision, ultimately, will be based on whether the tool has demonstrated its value and cost-effectiveness throughout the trial period. Feedback by both local and national stakeholders will be taken at different stages to gather wider input and explore broader applicability in the future.
Local adaptation and compliance: Given that the technology originated outside the UK, it requires thorough adaptation to align with national policing standards, legal frameworks, and evidential requirements – a process that remains ongoing.
This adaptation process is essential to ensure that the technology’s outputs are both legally robust and compliant with UK policing and evidence standards. For example, some data reaches UK policing in formats different from those Söze has previously handled – every bank uses a bespoke template, and even global social media platforms often have local variations. Adapting to challenges like this, the pilot phase continues to evolve close collaboration with key stakeholders and partners to assess the tool’s capacity to withstand legal scrutiny within UK courts. Successfully integrating technologies from overseas into UK policing requires not only technical adaptation but also a comprehensive understanding and response to local practices, legal considerations, ethical frameworks and public confidence.
A future built on innovation and collaboration: The future of policing lies not just in adopting new technologies but in doing so thoughtfully, ethically, and collaboratively. The successes of the Dutch NPAI Lab and the Söze platform scaling out of Australia are just a few examples of projects the UK can take note of and learn from. By integrating lessons around the world while addressing local needs, UK policing can evolve into a more adaptive, responsible, and effective force for the public.
