Book publishers, music companies, and animation creators have requested amendments to the AI development bill in Russia. They want neural network training on protected works to occur under license – with permission and compensation for authors.
Currently, the document provides for a different mechanism: the copyright holder can prohibit the use of their content for model training. However, market participants believe this is insufficient. In their opinion, developers can train AI on available texts, music, voices, and images without a separate agreement with the author.
The industry's main fear is that a neural network will first study an artist's style, song lyrics, or voice, and then be able to imitate them without paying the creators. Associations warn that such an approach will reduce the incentive to produce new Russian content and make the country a convenient platform for free training of large models.
The bill itself primarily concerns large fundamental models with more than 1 billion parameters. Copyright issues and responsibility for AI's results are largely intended to be addressed in future subordinate acts. The government promises to continue discussions with the industry, but the media market demands to be included in the work on amendments by the second reading.
Read more on the topic:
- From Data Centers to Laws: Ministry of Digital Development Prepares to Control the Entire AI Chain
- AI Law "Lost its Teeth": Labeling Canceled, Copyright and Liability Postponed
- AI under the law: State Duma prepares phased regulation of neural networks