Generative AI is revolutionizing all areas of life — including crime. "Deepfakes have become part of our daily lives and will gain even more significance in the future. Taking a systematic look at one of the most critical cases — deepfakes in law enforcement — from legal, philosophical, and technical perspectives is foundational work for the future," says Prof. Dr. Niklas Kühl, Chair of Business & Information Systems Engineering & Human-Centric Artificial Intelligence at the University of Bayreuth. "Today, almost anyone can fake virtually anything and anyone with relative ease. This presents opportunities but also significant risks — just think of voice-cloning scams like the 'grandparent scam.' But what if law enforcement agencies themselves were to use such voice clones to solve crimes?" asks Prof. Dr. Lena Kästner, Chair of Philosophy, Computer Science & AI at the University of Bayreuth. Prof. Dr. Christian Rückert, Chair of Criminal Law, Criminal Procedure & IT Criminal Law at the University of Bayreuth, explains: "Deepfakes pose significant challenges for law enforcement and criminal courts. For example, how should courts handle the likely objection from a defendant that the video evidence does not depict them but is instead a deepfake? At the same time, AI-generated deepfakes also offer great potential for law enforcement. Therefore, we are investigating whether and how law enforcement agencies can use deepfakes for undercover operations."
This project is financially supported by bidt. Its new research focus addresses the question of when and how much trust is warranted regarding generative AI — whether in its creation, interaction, or evaluation of AI-generated products. Beginning in 2025, bidt will fund 10 research projects on issues related to trust and AI. The University of Bayreuth’s project "For the Greater Good? Deepfakes in Law Enforcement" will receive nearly one million euros (€992,400).
“In this project, we are examining the precise circumstances under which the use of deepfakes in law enforcement may be justified and its implications for other areas of society. We are bringing together experts in computer science, philosophy, and law to develop guidelines for the use and identification of deepfakes,” explains Lena Kästner.
Without a thorough investigation, there is a risk that deepfakes will be employed without a legal basis conforming to human rights, the German Basic Law, or EU law. Conversely, law enforcement agencies should, where legally and ethically justifiable, be able to leverage the potential of this technology within the limits of the rule of law to ensure effective law enforcement. Kästner emphasizes: “It is important to curb criminal, deceptive, or manipulative practices involving deepfakes. At the same time, it cannot be the case that law enforcement agencies are unable to counteract criminal activities because they are not allowed to use state-of-the-art technologies. A legally sound balance must be found to address these seemingly conflicting demands.”
Over the next three years, Kühl, Rückert, and Kästner will collaborate with prosecutors from the Central Cybercrime Unit of North Rhine-Westphalia (ZAC NRW) to investigate under which circumstances and to what extent the use of deepfakes is socially acceptable.
