Policing Artificial Intelligence: Will AI Enable Criminals To Do More Damage Than Ever Before?

AI. It’s been the hottest topic in technology news ever since OpenAI’s ChatGPT took the internet by storm back in late 2022. Nvidia’s stock price is soaring, Google’s Bard is closing the gap, and the Sony World Photography Awards were won by Boris Eldagsen using Dalle-2’s image generation. Only after the flurry of daily limericks and Van Gogh style puppy paintings began to die down, did people start to ask the important questions. Questions like: How do we stop students from cheating on their coursework? Do millions of office workers need to start brushing up their CVs? How long before we’re all bowing down to our robot overlords? 

‍

With all the controversies about offensive generated responses, and copyright infringement, and AI hallucinations, the media has had little time to think about more traditional ways in which AI might enable public harm: by enabling criminals to do more damage than ever before. 

‍

What exactly will the AI revolution mean for law enforcement? Let’s take a look at the challenges and opportunities it presents.

‍

‍Challenges: Smarter Crime

Artificial Intelligence is a broad field. Put simply, it means using computer processing power, datasets and algorithms to perform tasks that previously required human intelligence. Given many of the public AI tools are multipurpose, it’s hard to imagine any type of crime that might not be impacted by AI in some way. There are certain areas, however, in which AI will enable incredibly novel criminal strategies:

‍

• Deepfakes - By using deep learning to produce images and videos and audio, criminals will have access to incredibly compelling tools for fraud. Take the example of the UK energy firm that was duped into transferring €220,000 to fraudsters by an AI-generated audio clip of the CEO’s voice. Social engineering will become a whole lot easier with AI tools providing unbelievably impossibly persuasive digital props ideal for tailored phishing attacks.. And it’s not just fraudsters that can use deepfakes. Inevitably, image generation software will be turned to creating illegal pornography and CSAM. Revenge porn has been a tricky enough problem for law enforcement to adapt to while the images and videos have been genuine. Just imagine when swaths of imitative images will be just a few clicks away for would-be perpetrators. For CSAM the forecast is even bleaker. Can you imagine an image-generation tool trained on a database of CSAM? Not only would this see the global proliferation of CSAM exponentially grow but it would become a major barrier to global attempts to identify victims too. Authorities in the UK have already reported finding suspected AI generated CSAM online earlier this year. Fortunately, startups developing digital provenance technology, such as ForceField in the US, are already emerging in this space.

• Password Cracking - While AI might help criminals use social engineering to defraud their victims, simply stealing what they’re after will also become easier with AI. By training AI models on password datasets, criminals can create tools easily capable of cracking passwords - yes even passwords with a symbol at the end. In order for our online accounts to remain safe, users will need to use increasingly lengthy and sophisticated passwords, in addition to more robust means like multi-factor authentication and biometric access.

• Automated Terrorist Attacks - With every Google traffic light Captcha you solve, AI driving gets a tiny bit closer. The stock market seems to think so, in any case, given Tesla’s record-breaking stock price. But it’s not for nothing that in every sci-fi depiction of self-driving cars, from Minority Report to I Robot, it’s the danger this technology might pose that is highlighted. Sadly, vehicle ramming attacks have become more prevalent than ever over the past decade, with the 2017 Barcelona attack perhaps being the most infamous. As AI-powered automotive technology makes our lives more convenient, it will also provide opportunities for remotely controlled terrorist attacks.

‍

Opportunities: Automating Police Work

‍

Luckily for law enforcement agencies across the globe, AI is a blade that cuts both ways. For every opportunity it creates for criminals, there is an opposite opportunity for law enforcement. By leveraging AI technology, huge strides can be made in fighting crime more efficiently and more effectively. Here are some of the most exciting areas:

‍

• Robotics - Drone technology has rapidly advanced over the past decade. For law enforcement the possible use cases are almost endless: crime scene photography and 3D LIDAR scanning, mobile or fixed surveillance, bomb disposal, remote patrolling, area searches, and much more. As AI evolves, drone systems will be able to deploy in an increasingly independent manner and automate tasks, such as surveillance, that currently carry heavy staffing burdens.

• Intelligent Software - Smarter hardware will no doubt have a big role to play for crime fighters in the future. But smarter software is essential right now. Patrol officers, detectives and analysts are already being overwhelmed by the sheer quantities of data they are dealing with on a daily basis. From CCTV, and body camera, and even ring doorbell footage, to open source intelligence, smartphone messages, and internet search histories, the vast swaths of digital data can be collected for almost every case reported has ballooned investigation workloads to unmanageable levels. Investigators simply do not have time to analyze it all and opportunities are being missed. AI software can automate a lot of the time-consuming routine tasks police officers carry out and connect dots that might otherwise be overlooked.

• Closing Cold cases - DNA advances have seen hundreds of high profile cold cases being solved by police over the past few decades - often by feeding suspect DNA into ancestry matching products. Quite possibly, AI advances could have a similar impact. Using AI to analyze and synthesis the huge collections of evidence these cases generate, is one way they’ll be able to help but there are many other technical developments that AI will enable too. For instance, researchers at Northumbria University, Newcastle  have employed machine learning to advance gunshot residue analysis in a manner that bypasses bottlenecks in current forensic procedures.

‍

The good news for law enforcement is that their technological capabilities are far higher than even the most successful of criminal organizations. Training algorithms on large datasets is not a cheap thing to do (for now) and, with innovative technology companies and enormous budgets at hand, government-funded police agencies are well placed to acquire phenomenally powerful crime fighting tools.

‍

The main challenge for law enforcement, therefore, is to adapt to the changing technological landscape with the same agility with which criminals will adapt. It’s commonly held business wisdom that it’s more difficult for large organizations to innovate than startups: leaders are often incentivized to preserve the status quo, internal politics hamper alignment, and budgets get allocated elsewhere. Criminals organizations will face none of these hurdles, however, and with the pace of technological change accelerating exponentially over the past decade, crime fighters will fall behind if they do not act swiftly.

‍