On February 23rd 2023, Joseph Cox, a journalist at Vice, wanted to prove how AI, and even the cheapest one, can help breaking into a highly secured account. In order to do that, he chose his own bank account and managed to log into it. At first it may not seem concerning, but it is clear that voice authentication in accessing bank accounts creates a perfect ground for robbers. What should be striking in that story is the ease with which it can be done. Logging into important accounts requires multiple identification checks to be sure someone is not misusing yours. Identity fraud is recognized as a crime in France and Spain but how does it work when AI gets involved ? In this case using AI to log into your own bank account may seem entertaining but it shows how effortless it is to commit a crime with the help of AI. Most importantly, are states able to use this tool to prevent crimes as well.
The ROXANNE platform is a tangible example of a project combining criminal network and speaker data analysis. The project uses various speech processing methods such as time-relation analysis or meta-information as link analysis, which track and uncover criminal identities as well as identifying criminal networks. In collaboration with the EU and Interpol, ROXANNE aims to enhance investigation capacities especially in the case of large crimes whilst remaining in line with legal and ethical frameworks.
One of the regulations on AI is the EU’s AI Act, which shows us the need to regulate it and prevent its “malicious uses or abuses”. This regulation aims to prove that within European Union, “technology is used in a way that is safe and compliant with the law, including the respect of fundamental rights”. According to the act, AI systems are classified in different categories depending on their risk of causing harm. Law enforcement is one of the fields listed that is allowed to operate with high-risk AI systems. However, the field of activity is framed by strict requirements. As a result, even in law enforcement, the use of remote biometric identification is prohibited unless victims of a crime are searched, a serious crime or a life threatening situation is at stake.
The use of personal data in the case of law enforcement raises debates inside the EU. On the one hand, it is believed that giving police more freedom will facilitate bureaucratic work and reduce crime rates. On the other hand, scepticism about the automation of policing raises the attention on a more restrictive approach to AI’s use in law enforcement.
A very recent example is the use of a surveillance software by the police of two states in Germany that has been declared unconstitutional by the German Federal Constitutional Court. In the case of the state of Hessen, the software was used for predictive policing that consists of data mining in order to profile potential criminals before a crime has been committed.
Especially in crowded cities where criminality rates can be quite high, integrating AI is a must. It can save officers a lot of time from paperwork as well as help them “focus on high-value, rewarding tasks that involve complex reasoning”. However, there are many cases in which personal data was misused and led to unjust prosecutions. Therefore, the issue is less about law enforcement using AI or not, and more about how data is manipulated, stored and exchanged.
Since our modern societies only have the opportunity to use AI as a tool, its use by the police may seem abstract and intimidating. Will it be controlled precisely and respect ethical standards ? The different examples previously described offer us a sensible application of AI to the law enforcement environment. But the efforts of regulation at the European level, must continue meanwhile AI updates itself rapidly and becomes more widespread. Enhancing cooperation between states is one of AI’s supposed aims but it should not be put at risk by some pernicious ambitions. Next time you enter a new country through gates with facial recognition, think about how much information that country's government has got on you.
If you’re interested in this topic, we recommend you watch Person of Interest (2011), an American TV show that foresaw a fictional world in which AI surveillance was used to prevent crimes.
By Eda Üner and Pauline Royer
To go further, on our April file dedicated to artificial intelligence, you can read our Culture and International Relations sections' articles on our blog.
Follow MINERVIEWS
This newsletter is made thanks to the effort of Lyon II university and Auvergne Rhône-Alpes region.
