This article discusses the principles of and premises for liability for damage caused by AI systems. It applies to liability models based on the principles of risk and guilt. It indicates that different groups of entities, e.g. programmers, may be responsible for the creation of AI under the principle of guilt, while producers and merchants may put it into circulation under the principle of risk. The liability of AI system users should be tempered and based on the principle of guilt. This article includes a critical view of the AI Act and the relevant directives. It points out that effective liability for damage should be related to the level of harm caused (harm to a person, human death) and not dependent on whether it was inflicted by a high-risk system or any other AI system.