Answers are grounded in this paper and cite their sources.
Artificial intelligence can be a very useful tool for creating new databases. However, such use raises many complicated legal issues, ranging from determining the protectability of such databases to the risks that users face when using them. A separate group of problems is related to the specific issues of distributing large language models. First of all, the operation of artificial intelligence is not transparent, that creates serious legal risks both for the users of the databases being created and for the developer themselves. Next, attention should be paid to the difficulty of proving the protectability of databases created with the help of artificial intelligence — and it requires a change in the approach to assessing these objects and the costs of their creation. This applies both to the field of copyright law (assessing the “selection and arrangement” of the materials that make up the database) and to the related rights of the database producer (determining the substantiality of investments in the creation of the database, since the activity of artificial intelligence in fulfilling an individual request usually does not require significant costs in itself). It appears that no substantial legislative changes are required. The focus could be on adjusting the content of existing regulations through an interpretation that aligns with new social and technological conditions. Substantial risks arise when using databases created with help of artificial intelligence. Here, it is worth noting the nonobviousness for the database user of the absence of an author, and the high risks of infringing third party rights. Serious difficulties are caused by model hallucinations. It is important to understand that the occurrence of hallucinations does not mean errors in the implementation of a specific language model, but is a natural feature of modern large language models as a whole. They can be minimised, but not eliminated entirely. At the same time, measures taken to reduce the number of model hallucinations often lead to a decrease in the completeness of the response. In both cases, the quality of the response generated by the model suffers. However, all issues mentioned seem solvable, although they require the attention of legislature, judicial practice, and, of course, users themselves. In this regard, a contract becomes an important tool, through which the parties can both resolve issues related to securing rights to the intellectual property created and allocate responsibility (and the associated risks) for the content of the material prepared using artificial intelligence.
No takes yet. Share an insight, caveat, or question.
Виталий Олегович Калятин (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: