Abstract: The subject of the research is the influence of artificial intelligence on the liability of the launching state in international space law. The main focus is on the situation where a space object operates with a high degree of autonomy, and the reasons for its decisions cannot be easily explained after an incident. This issue is known as the "black box" of artificial intelligence. The article examines how this opacity affects the establishment of causation, proof of fault, and the distribution of evidentiary risk. The responsibility for harm caused in space, where a fault-based liability regime applies, is analyzed separately. The research also addresses issues of human control, auditing, logging, data verification, and national licensing of AI-enabled space systems. The work demonstrates that the autonomy of the algorithm does not eliminate the connection between the space object, the operator, and the launching state. The methodological framework of the research incorporates legal formalism, systemic analysis, comparative law, and teleological methods. These are employed to analyze the norms of space law, liability regimes, and contemporary approaches to regulating artificial intelligence. The scientific novelty lies in examining the "black box" problem not as a technical justification for exemption from liability but as a legally significant factor in proving cases. The article proposes the use of a standard of due algorithmic diligence. This includes risk assessment, model verification, establishing limits of autonomy, data retention, decision logging, and the possibility of subsequent auditing. It concludes that artificial intelligence should not be recognized as an independent subject of international liability. Fault should be assessed through the actions and omissions of the state, the operator, and other participants in space activities. If the state or the operator has not ensured the verifiability of the system, the risk of non-proving should not be shifted to the injured party. In such cases, a relaxation of the burden of proof or the application of a rebuttable presumption of fault is possible. At the same time, the liability of the launching state should have reasonable limits and take into account the predictability of risk, the nature of the damage, and the distribution of functions among mission participants.

Keywords: international space law, artificial intelligence, black box, launching State, space object, autonomous systems, causal link, evidentiary risk, algorithmic due diligence, algorithmic transparency