European Police Congress: Investigators need AI and small data

Contents

The use of “big data” in connection with police investigations is controversial – “small data”, on the other hand, is new to the debate. However, the problems seem to be basically the same and police officers and data protectionists often disagree. This was demonstrated during a debate on the use of technology in police forces at the 28th European Police Congress in Berlin.

Carolina Stindt from the management consultancy PwC presented a study based on interviews and discussions with police officers. According to the study, the use of technology in traditional areas such as video surveillance, border control and case processing is comparatively well established. However, newer technologies, in particular AI-supported technologies, are “often only used tentatively and on a pilot basis” in individual federal states and police authorities in Germany. According to Stindt, the framework conditions are often “pre-digital” and the vast majority of interviewees said “that these technologies would be particularly important in order to make data collection and evaluation more efficient and, in particular, to better recognize patterns”. However, Stindt also pointed out that data-based decisions in particular still meet with great skepticism among the population.

Andreas Stenger, President of the LKA Baden-Württemberg, had frank words. He is critical of the status quo of equipment in the police forces: the IT landscape is heterogeneous, some federal states are advanced, others are not, some technology is outdated, the infrastructure is dilapidated, fiber optics are problematic, automation processes are sometimes difficult to implement because the infrastructure is lacking. The main problem for him as a practitioner is the slow implementation of the Saarbrücken Agenda. With a little sideswipe at three of the discussion participants (including the moderator), who work at consulting firm PwC, he suggested “more governance and less external consulting”.

“Great data” cannot be used due to a lack of regulations

For Gerald Eder, the technical director of “Programme 20” for the digitalization of the police, the heterogeneous initial situation in a federal state is the biggest challenge. Even the question of which officer has access to which information is a huge coordination process. According to Eder, we are in a dilemma: we currently have some “great data”, but it is not available because the relevant regulations do not yet exist. He cited the Magdeburg attacker as an example – The investigators involved in the case had only seen data from Saxony-Anhalt, but not the amount of data collected on the perpetrator nationwide.

Publicist Anke Domscheit-Berg, former Left Party MP, pointed out that it is always about balance and referred to the Federal Constitutional Court. This requires consideration: Is a measure even demonstrably suitable for solving a problem? And if it is suitable, but means an encroachment on fundamental rights: Can it be replaced by another measure? And is the measure appropriate, i.e., are the “side effects” proportionally the expected benefits? “This often works very badly in practice,” Domscheit-Berg stated.

What's more, politicians keep repeating mistakes. For example, the issue of sexualized violence against children. Chat control has proven to be unsuitable here, for example due to a high error rate caused by false positive hits from sexting among young people. Domscheit-Berg called for more prevention instead. At a hearing in the Digital Committee, a public prosecutor from North Rhine-Westphalia reported “that they also have evidence lying around for three years before it is viewed for the first time.” Why, asked Domscheit-Berg, are we talking about AI “if they simply don't have enough people?”

The supreme discipline of security

According to Peter Beuth, PwC, the police need more data and must evaluate existing data better, especially for emergency response, the supreme discipline. Above all, they require technology to be able to link data – this is only possible with AI and not manually, as there are not enough officers for this. You have to link existing data, small data. However, not only small data, but also big data, for example for averting danger at major events where crowds of people are on the move: “Big data will also have to exist for successful police work in the future,” stated Beuth.

Andreas Stenger was convinced that there were already essential basic requirements for the safe use of modern technology, but of course, there were also obstacles due to federalism: for example, some state commissioners for data protection would assess the AI Act differently, and he would like to see harmonization across Germany. Stenger cited facial recognition software as an example: this is used by the BKA and the LKA Baden-Württemberg. It takes three to five years to train humans to recognize people; with AI, it has been possible to reduce the error rate from 30 to less than one percent. It is also quicker, even if a forensic expert ultimately makes the decision and represents it in court.

Controversial discussion

The different views became clear in the debate. One of the topics was the alleged RAF member Daniela Klette, who was tracked down by journalists using facial recognition software. Do the police need such technology? Are they allowed to use it? Domscheit-Berg pointed out that some services are illegal in Europe, people do illegal things, but she expects institutions of power to refrain from doing so. The police seem to be in a dilemma when it comes to such issues: According to Stenger, malice was poured on the police after Klette's arrest. He would have liked a clarification: “For historical reasons, we do not have a Reich Security Main Office.”

The 28th European Police Congress took place in Berlin from May 20 to 21, 2025. Over 2,300 participants and 230 speakers gathered information in numerous lectures, specialist forums and discussion panels.

(mki)