Confused in neural networks: why the new AI bill has attracted sharp criticism
While ordinary users are binge-eating AI content, not distinguishing real video characters from virtual ones, and listening to music generated by neural networks, representatives of the creative industries are sounding the alarm. On Tuesday evening, the State Duma approved in the first reading the draft law "On supporting the development of artificial intelligence technologies in the Russian Federation." One of its clauses allows the use of works published on the Internet for AI training without the consent of the authors and payment of remuneration to them. About why the current version of the bill may deprive people of art of the last piece of bread — in the material "Izvestia".
Advanced technology or a return to piracy
The new document, which the State Duma considers as a step towards the development of domestic technologies, has already provoked a sharp reaction from writers, publishers, musicians, producers and directors. And it's not that anyone opposes new technologies. The question is whether the training of domestic AI programs should take place at the expense of the intellectual work of real people. And for free. This is exactly the conclusion that suggests itself after reading the text of the bill.
In fact, the document is not devoted to all artificial intelligence, but to the largest generative models like YandexGPT, GigaChat, ChatGPT, Gemini and DeepSeek. It is for them that it is proposed to create special rules of the game.
The most controversial rule allows the use of copyrighted works to train neural networks without the consent of the copyright holder. Although the current civil legislation proceeds from the exact opposite principle.
In fact, the bill proposes to make an exception for the creators of AI, which is not available for any other market participant. The developers call the opt-out mechanism a compromise — the author's right to independently prohibit the use of his works for training neural networks. However, in practice, this approach shifts all responsibility onto the copyright holder himself. He will have to keep track of where his content is published, search for copies of it, including on pirated resources, send notifications and monitor the implementation of the ban.
The situation with pirated content is even more strange. If a work is illegally posted on a third-party site and is included in the training sample, the copyright holder often does not even know about it. It turns out to be a paradox: first, the state has been fighting Internet piracy for years, and then it offers a model in which illegally posted content can become a raw material for commercial artificial intelligence systems.
Evgeny Kapyev, General Director of EKSMO Publishing House, shares these concerns. In his opinion, the main disadvantage of the bill is the use of intellectual labor results without the consent of copyright holders.
— It is impossible to protect it with technical means, because there are pirated libraries. The second story is that if legitimate copies are bought, it means that an electronic copy can also be used for training, but this also contradicts the civil code," Kapyev said in a conversation with Izvestia.
Director Nikita Vysotsky is also concerned about the issue of fair remuneration for the authors. He is convinced that the development of AI should not take place at the expense of the free use of other people's creativity.
— In Europe, there were so—called pirate parties that also proposed to abolish intellectual property altogether - supposedly it would be easier for everyone and everyone would develop. They do not understand that if people lose the right to what they have created, they will not give it back in the future, they will not make it public. They will look for ways to avoid being robbed. This will slow down the development. As far as I understand, artificial intelligence cannot create anything on its own — it compiles from what has already been created before it. Pay the money. You won't be providing the services of this artificial intelligence for free, will you? — the director told Izvestia.
Music producer Joseph Prigozhin also sharply criticized.
— There is such a thing as intellectual property: it must be protected. Robots are also trained by humans. I understand that I don't want to incur costs for experiments, but let it be a certain amount, conditionally agreed. If it gets distributed, the author receives a reward. It won't work purely without investment and without costs. Many authors have dedicated their lives to writing, and maybe someone has even had one work in their entire life. And taking away the last piece of bread from him is not quite right," said Prigozhin.
What do they say about the bill in the State Duma
The deputies who are involved in drafting the bill emphasize that the initiative does not abolish copyright and does not give AI the right to freely use other people's songs, films, texts, scripts, voices or images. However, the concerns of the creative environment in Parliament are considered justified. We cannot allow the development of AI to take place through the free use of human labor, Marina Kim, first deputy chairman of the State Duma Committee on Information Policy, Information Technologies and Communications, told Izvestia. In the future, the deputies will finalize the bill.
— AI can be a tool, but the author and copyright holder remains a human being. When further refining legislation, it is important to discuss separately the mechanisms for protecting copyrighted content: the consent of the copyright holder, transparency of the use of works, fair remuneration and responsibility for illegal imitation of a voice, image or works," the parliamentarian said.
Earlier, Dmitry Grigorenko, Deputy Prime Minister and Chief of Staff of the Government of the Russian Federation, said that authors would have the right to protect their work from AI. The deputy Prime Minister's office explained to Izvestia that the bill introduces a mechanism that allows content creators to independently prohibit the use of their works to train AI models.
"If the copyright holder has sewn up special hidden tags (markup that is read by algorithms) in the code of a web page where copyrighted works are published, such materials are prohibited from being used when training artificial intelligence models,— Grigorenko's office said.
Thus, the authors have a legally significant tool to protect their content. If the developer of the AI model has ignored the ban, the copyright holder still has the opportunity to file a claim.
It is important that the bill is specifically about using copyright objects to train models, and not about copying works — when a neural network reproduces a song, movie, or text so close to the original that it may violate copyright, the press service of ANO Digital Platforms, where experts are actively involved, explained to Izvestia. We discussed the initiative.
Who supported the new bill
However, not everyone sees the danger in the new bill. Writer Alexander Tsypkin considers the discussion of the document to be "the fight against windmills."
— The only thing I would like to protect in this case is the last name. If my last name is used in an advertisement for some text and it says: "I was trained on the texts of Alexander Tsypkin," then, probably, I will demand some kind of financial proceedings. Otherwise, I want to ban artificial intelligence from learning from my texts.… How do you imagine it? It's impossible. Your face and last name can be protected. And everything else works rather poorly. We are entering the era of a new round in which all the old forms, including legal ones, lose their significance," the writer told Izvestia.
At the same time, director Vladimir Kott is convinced that neural network training is inherently closer to the process of human knowledge acquisition than to the commercial use of works.
— When a child comes to school to study, he reads not only classics, but also modern literature. Training does not involve a commercial approach. We don't ask the authors if they can be read and studied. It's a matter of context. I think the authors should be happy that their content is used for teaching," the director said in a conversation with Izvestia.
Today, millions of users already receive news and answers to socially significant questions not through media sites, but through AI assistants. The bill actually legalizes the possibility of using journalistic materials to generate such responses, but says practically nothing about the responsibility of the owners of these services.
If a newspaper publishes false information, the editorial board is responsible for the consequences. If a TV channel violates the law, there is a clear system of responsibility. And who will be responsible if the neural network gives out false information about human health, distorts the news agenda, or simply "hallucinates" by inventing non-existent facts?
So far, there is no answer to this question in the bill.
The development of artificial intelligence really requires a modern legal framework. But creating it by weakening the institution of intellectual property is a risky path. Without authors, there will be no books, music, films, journalism, or scientific research. And without new high-quality content, artificial intelligence itself will eventually cease to develop.
Yvette Nevinnaya, Victoria Sytnik, Maxim Bazanov, Daria Emikeeva, Ekaterina Sidorovich
Переведено сервисом «Яндекс Переводчик»
