More cases: Child abuse with AI: Cyber ​​investigators face new problems

In the fight against child pornography, material created with artificial intelligence (AI) presents investigators with new challenges. “So far there is hardly any case law on the subject,” said Thorsten Nees, first public prosecutor at the Baden-Württemberg Cybercrime Center (CCZ) at the Karlsruhe Public Prosecutor’s Office. The center processes a double-digit number of cases and secured one of the first convictions in the area nationwide.

Fictitious child pornographic content is also child pornography within the meaning of the Criminal Code. But what matters is whether the material can be clearly recognized as artificial from the perspective of an average viewer at first glance, explained senior public prosecutor Mirko Heim from the CCZ, citing manga as an example. Then there would only be a penalty for distributing it and making it accessible. If this is no longer so easy to determine, possession and the so-called acquisition of possession can also be punished with several years in prison. The problem with this: “What is realistic is a question of interpretation,” explained Nees. This always entails legal uncertainties.

During an analysis, a research team from the University of Passau came to the conclusion that operators of AI tools could also be prosecuted under certain circumstances - for example for aiding and abetting.

Perpetrators use freely available tools

Cases of AI-generated child and youth pornography are increasing significantly, the experts said. “This is also because it no longer requires any special IT expertise to create such content using freely accessible AI software,” explained Nees. The perpetrators regularly used AI tools freely available on the Internet. The CCZ did not provide any specific figures.

According to investigators, the perpetrators are primarily interested in satisfying their sexual interests. Sometimes such content is used to harm the person affected or the person depicted.

In order to recognize that an image was artificially created, input commands (prompts) and proof of origin are particularly helpful. In general, this is becoming increasingly difficult, admitted Nees. If content is anonymous, encrypted or hosted abroad, this can make identifying the perpetrator more difficult.

Why is the distinction AI or not important?

According to the information, whether it is real or AI-generated child pornography is important not only for the actions of the investigating authorities or with regard to criminal liability. When it comes to sentencing, it could also play a role that no real child was abused.

However, Heim emphasized that children whose likenesses were used felt abused and ashamed, similar to a real abuse victim. On the other hand, there is a risk that a perpetrator will want more over time and will either switch to real child pornography or attack a child in real life. “Every AI-generated image can lower the inhibition threshold,” said Heim.

The Federal Commissioner against Sexual Abuse of Children and Adolescents, Kerstin Claus, made a similar statement at the beginning of the year: “Such recordings are extremely problematic because they not only depict sexual violence, but reproduce and normalize it.”