Violation of Legal Interests by AI Systems in the Light of the Function of Criminal Law – AI and General Trust

• Applicant: Nicolai Preetz
• Subject Area: Criminal Law
• Term: since 2025
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 565687773

Modern society faces the challenge of integrating AI into almost all areas of life. The idea behind the thesis stems from the realization that the phenomenon of violations of legal interests by AI systems is a still unsolved problem that can be explored in depth from the perspective of the function of criminal law and can be a testing ground for a new theory of criminal law based on the mechanism of social „Generalvertrauen“, which can be roughly translated as „general trust“. The aim of the thesis is therefore to show how violations of legal interests by AI systems affect the function of criminal law and how the criminal law system as a social system can react to this. In doing so, the thesis aims to contribute to the legal-theoretical and legal-sociological discussion on the function of criminal law, to uncover systematic gaps in criminal liability and to offer possible solutions in the communication of the criminal law system for dealing with this phenomenon. The function is the central point of reference and the first question. A theory of criminal law was developed around this question, the “Theorie vom Generalvertrauen”, which draws on constructivism and Luhmann's systems theory as a theoretical foundation and explains criminal law in its social context from a new perspective: as a guarantor of general trust. General trust is a mechanism with which humans can cope with the complexity of the modern world by reducing expected possibilities. In conjunction with the phenomenon under investigation as the first application of the theory, the main thesis follows: criminal law serves to maintain and constitute general trust. Violations of legal interests by AI systems potentially destroy general trust. The criminal justice system can respond communicatively to this – by attributing responsibility to humans, by communicating about non-punishment and, in the future, possibly by attributing responsibility to AI systems. It was shown that, in most cases, the criminal law system is able to assign responsibility in a socially connectable manner to the people behind the AI systems. Systematic gaps were identified due to evidentiary challenges in the context of criminal liability of the manufacturer in the context of breach of duty, and it was also shown that principles of objective attribution are not consistently observed in the case of criminal liability of the user. In addition, criteria were identified under which AI systems could be criminally liable in the framework of societal constructs in the future. Solutions for the remaining gaps in criminal liability were identified. The perspective of connectable communication about non- punishment draws attention to the communicative achievements of the criminal justice system beyond traditional punishment. It shows how nonpunishment also serves the function and how other social systems maintain functionally equivalent general trust.

• DFG Programme: Publication Grants