AmCham Position Paper on the Criminal Liability Framework for Artificial Intelligence

In this position paper, AmCham analyzes the new criminal offense of causing danger through the use of artificial intelligence systems, introduced by amendments to the Croatian Criminal Code in October 2025. While supporting effective protection of citizens and society from the harmful and irresponsible use of artificial intelligence, AmCham warns that the existing approach creates legal and regulatory uncertainty for the business community.

The position paper particularly highlights the overly broad range of potentially liable persons, encompassing nearly all participants in the AI system lifecycle, as well as the imprecise definition of “causing danger.” Unclear boundaries of criminal liability, combined with severe penalties, could negatively affect investment, innovation and the adoption of new technologies, particularly among startups, small and medium-sized enterprises and research organizations.

AmCham also emphasizes that European and international regulatory frameworks primarily focus on risk management, transparency, human oversight, and regulatory and civil liability mechanisms. Criminal law is generally used as a last resort and reserved for clearly defined and particularly harmful forms of AI misuse.

Accordingly, AmCham recommends linking criminal liability to specific breaches of safety and regulatory obligations prescribed by European and national legislation. It also proposes limiting criminal liability to clearly defined and particularly harmful forms of misuse, such as identity manipulation, serious fraud and threats to critical infrastructure, while applying the principles of proportionality and gradual escalation of liability.

The position paper is available here.