Opinion
The Double-Edged Sword of AI
By Lucy Toom-Smith, Altia
AI is moving faster than most of us can keep up with. In policing and investigations, that speed is exciting – and honestly, a bit unsettling. One week we’re celebrating a clever new way AI has helped detect crime or prevent fraud, and the next we’re reading about a system that’s made a mistake with very real consequences. The recent UK policing reform whitepaper made one point unmistakably clear: technology will shape the future of policing. But it must do so in a way that strengthens public trust rather than eroding it.
AI can analyse information at a scale no human team could ever match.
AI can flag unusual financial activity in seconds, sift through thousands of documents, or highlight patterns that would otherwise remain hidden. Working with investigators at Altia has shown me just how valuable that capability can be when used well. But it has also shown how quickly things can unravel when AI is treated as infallible.
We’ve already seen what happens when speed overtakes scrutiny. The mistaken justification behind the ban on Maccabi Tel Aviv fans at an Aston Villa match is just one example of how AI generated errors can slip quietly into real world decisions. These mistakes aren’t dramatic Hollywood style failures – they’re subtle, plausible, and easy to miss unless someone is actively checking the logic behind them.
Fraudsters, of course, are embracing AI with none of the caution. Fake accounts, deepfakes, and automated phishing campaigns are evolving at a pace that would have been unthinkable even a year ago. They don’t need specialist expertise; they just need access to the same tools everyone else has. Police and investigative teams are facing threats that adapt as quickly as the technology itself.
The conversation can’t simply be about “using more AI”. It has to be about using it responsibly.
The whitepaper’s emphasis on transparency and accountability isn’t a bureaucratic hurdle, but a practical necessity. If investigators are expected to rely on AI generated insights, they must understand how those insights were produced. They need to see the evidence, reasoning, and level of confidence behind each conclusion. Without that, AI risks becoming a black box that introduces new vulnerabilities rather than solving existing ones.
At Altia, this is the challenge we are working to address. Our focus is on explainable AI – tools that don’t just present an answer, but show their working, and require human approval. Instead of a single output, investigators see the data points considered, the logic applied, the confidence score attached, and any known weaknesses are flagged. This isn’t about slowing things down; it’s about ensuring that speed doesn’t come at the cost of accuracy or fairness.
I’ve seen first-hand how AI can genuinely enhance decision making, and how easily it can mislead when its outputs are taken at face value. The technology is powerful, but it is not neutral, and it certainly is not always right. Human judgment remains essential, not optional.
If policing is to benefit from AI, explainability must be treated as a core requirement. It’s the difference between a tool that supports investigators and one that undermines them. It is also the difference between public confidence and public suspicion.
The path forward is clear: adopt AI, but do it carefully. Demand transparency. Build systems that can be questioned, audited, and understood. And ensure that technology enhances – rather than replaces – the expertise of the people who use it. Doing so will ensure that the technology genuinely strengthens investigative work and supports the wider reforms the sector is striving to achieve.
