The government proposed allowing the use of copyrighted materials for training AI - Vedomosti

The Russian government has proposed allowing development companies to use copyrighted articles, books, films, images, etc. to train AI without obtaining permission from their copyright holders. This provision was included in the bill on artificial intelligence, which was prepared by the Cabinet of Ministers, writes the Vedomosti newspaper.

It is assumed that it will be possible to use works without the consent of the author if users of the service ultimately do not see their content.

In this case, the copyright for the answer of the AI ​​model will be received by the user who entered the prompt (a request to the neural network), processed the result and made his creative contribution. This rule will apply regardless of what data the model was trained on.

Vedomosti's interlocutor, close to the government apparatus, clarifies that we are primarily talking about scientific publications, educational materials and archival documents, access to which is often closed to developers of AI solutions. It will not be possible to use personal data, information about taxpayers, private correspondence and other sensitive data to train models.

The fact that the government is developing a bill to regulate AI became known in early February of this year. As of today, there is no final version yet, Vedomosti was told by the office of Deputy Prime Minister Dmitry Grigorenko, who oversees the industry.

Developers of AI solutions regularly face copyright issues. For example, Meta and Anthropic were accused of violating them. Thus, Universal Music Group and other major record labels filed a lawsuit against the latter, accusing the company of using copyrighted songs and lyrics to train the chatbot Claude.