- Статьи
- Society
- Turnout figure: AI-generated images of people are proposed to be banned in election campaigning
Turnout figure: AI-generated images of people are proposed to be banned in election campaigning
The State Duma Committee on State Building and Legislation has proposed to the lower house of parliament to support in the first reading the draft law on amendments to the legislation on elections. Among other things, it prohibits the use of images of people created using artificial intelligence technologies in election campaigning. Izvestia investigated how technically possible it is to control and detect the use of artificial intelligence in campaigning, as well as how such restrictions may affect the work of political headquarters.
Campaign progress
The use of images of people created by artificial intelligence during election campaigning can significantly distort the free expression of the will of voters, Dmitry Orlov, Director General of the Agency for Political and Economic Communications (APEC), said in a conversation with Izvestia. The situation is particularly sensitive when such technologies are used to recreate images of deceased political figures with strong symbolic influence. Historical figures like Lenin, Stalin, or Zhirinovsky, "animated" with the help of AI, can have a noticeable effect on the perception of the campaign and the behavior of the electorate.
In such cases, political forces resorting to such images can actually do without a full-fledged political program or meaningful appeals to voters. Instead, they use the emotional and historical capital of famous personalities, which is only indirectly related to their own activities and political positions.
"The law banning the use of AI—generated images of people in campaigning is aimed at a reasonable correction of the situation in the interests of the majority of voters," the expert noted. — Technically, it will not be easy to implement it, but the law may become a significant constraint already in the parliamentary campaign in 2026.
There are separate technical methods that make it possible to identify images created using artificial intelligence, but it is not yet possible to guarantee one hundred percent detection of such materials, including those used in propaganda, Lyudmila Gontar, director of the Aerodynamics project office of the Government Commission of the Russian Federation, lecturer at the Department of Applied AI at Rudn University, told Izvestia.
In some cases, Safe AI tools help. For example, if an image is generated in ChatGPT or DALL-E 3 (OpenAI), its origin can sometimes be established through the C2PA standard. However, when using systems such as Midjourney, Flux, or Stable Diffusion, especially with local generation, it is usually difficult to identify the source due to the lack of watermarks and metadata. If the metadata is intentionally deleted, it is almost impossible to establish the use of AI.
— Also, interesting trends are gradually emerging in the software market, such as AI or Not, Illuminarty. But they don't give an exact answer to the question either," she noted. The C2PA Manifest is a digitally signed block of data embedded in the file, which makes any change to the image violate its authenticity. Such a mechanism can be an effective tool for controlling the origin of content.
However, its implementation, according to her, requires changes in the regulatory framework, primarily in the field of digital signature regulation, as well as adaptation of the standard for use in the Safe AI system in Russia. In this regard, it is necessary to develop separate technical standards that allow verifying the origin of images and confirming their authenticity.
Continuation of the policy
The ban proposed by the deputies on the use of unreliable images and videos in election campaigning is a continuation of existing restrictions, Elena Gorodetskaya, PhD in Law, associate professor at the Department of Parliamentarism and Interparliamentary Cooperation at the RANEPA, told Izvestia. Its purpose is to ensure the genuine expression of the will of the voters. To do this, people should receive truthful information about candidates and form their political position freely and consciously.
According to her, government agencies and election commissions are required to ensure the accuracy of information, but it is more difficult to control this in the campaigning activities of candidates. Therefore, the legislation is trying to protect voters from manipulation, especially those used by candidates or other forces that do not have sufficient political "background" of their own and are trying to attract support by artificial methods.
Gorodetskaya noted that this would not create serious problems for election headquarters, because according to the law, images of the candidate and his family members can be used in campaigning. At the same time, the new ban will give law enforcement agencies additional tools to combat illegal campaigning on the Internet, for example, to remove fake images or videos from websites and social networks at the request of election commissions through Roskomnadzor.
There are technologies and detectors that make it possible to determine whether an image has been created using artificial intelligence. In her opinion, the supervisory authorities have the technical capabilities to identify and prevent such violations. In general, this measure, she believes, should protect voters from manipulation and help ensure the integrity and authenticity of elections.
Election campaigning is one of the most sensitive types of public communication, as it can influence the choice of voters and, as a result, the development of society, Olga Pozdnyakova, head of the Popular Front, told Izvestia. Analytics". In this area, there is a risk of using neural networks to create a distorted image of a candidate or a political association. Although AI image generation can be faster and cheaper than traditional methods, the right of citizens to receive reliable information must remain a priority.
From a technical point of view, cases of using neural networks to create text or visual content can be identified using the same artificial intelligence technologies.
— There are a number of services that are not perfect yet either, but they are constantly evolving. In most cases, these services distinguish real images from generated ones, but so far, high—quality images generated and especially modified by artists are not always defined as created by AI," the expert added.
Izvestia sent a request to the Supreme Court and the Ministry of Finance, but by the time the material was published, no responses had been received.
Переведено сервисом «Яндекс Переводчик»
