Artificial Intelligence has become one of the most influential technologies of the modern era. AI systems are increasingly used in healthcare, education, transportation, industry, finance, commerce, and public services. As these systems become more capable of analyzing information, generating recommendations, and performing tasks with varying degrees of autonomy, an important legal question emerges: who should bear civil liability when an AI system causes harm? Civil liability generally aims to provide legal remedies or compensation when a person suffers damage resulting from conduct, a defective product, or the breach of a legal or contractual obligation, depending on the applicable legal system. Applying traditional liability principles to AI, however, can be challenging because numerous actors may participate in designing, developing, supplying, deploying, and operating an AI system. When conventional technology causes harm, identifying the responsible human action may sometimes be relatively straightforward. In complex AI systems, however, an undesirable outcome may be influenced by training data, model design, software configuration, deployment decisions, user behavior, and the environment in which the system operates. Determining the precise source of failure can therefore become considerably more difficult. Automated driving technologies provide a useful example. If an AI-supported driving system contributes to a traffic accident, questions may arise regarding the responsibilities of the vehicle manufacturer, software developer, component supplier, vehicle owner, or user. There is no universal answer because liability depends on the cause of the accident, the applicable law, and the legal duties assigned to each party. Similar challenges appear in healthcare. AI systems can assist professionals in analyzing medical images or generating clinical recommendations. If an inaccurate output contributes to harm, determining liability may require an examination of the system developer's responsibilities, the healthcare institution's procedures, professional oversight, and the way the technology was actually used. The concept of fault is therefore central to many discussions of AI liability. Traditional liability rules may require proof that a person or organization failed to meet an applicable standard of care. Complex AI systems can make this difficult when injured individuals lack access to the technical information necessary to understand how a particular output was produced. Causation creates another challenge. A claimant may need to demonstrate a sufficient connection between the alleged fault and the resulting damage. In AI-related cases, multiple factors—including training data, model behavior, software settings, user decisions, and external conditions—may contribute to the final outcome. Product liability principles may also become relevant when AI is incorporated into products or services. If harm results from defective design, inadequate safety measures, or insufficient warnings and instructions, the applicable legal system may impose responsibilities on manufacturers or other economic operators. Another difficulty is the number of participants involved in the AI value chain. One company may develop a model, another may supply data or technical components, and another organization may integrate and deploy the system. Clear allocation of responsibilities is therefore important to prevent technological complexity from becoming an obstacle to effective compensation. Transparency and traceability can play an important role in addressing these challenges. Maintaining appropriate records of system development, testing, deployment, updates, and significant decisions can assist investigators and courts in identifying the causes of harmful outcomes.