Sick fantasies: AI is increasingly being used to create child pornography

The volume of images and videos of sexualized child abuse, as well as other extreme content created using artificial intelligence (AI), has reached record levels, experts have warned. Experts urge to update the laws on Internet security, as well as introduce strict rules for relevant companies. For more information about why neural networks are increasingly being used to create child pornography, what is the danger of this trend and how to deal with it, read the Izvestia article.

What is known about the records of prohibited content generation

The fact that the volume of generation of images and videos of sexually abused children, as well as other extreme content created using AI, has reached record levels, according to a report by the British non-profit organization Internet Watch Foundation (IWF).

It is the IWF that owns Europe's largest hotline for reporting such materials. According to the report, in 2025, the organization identified 8,029 realistic images of sexualized child abuse — 14% more than a year earlier. At the same time, analysts also checked more than 3 thousand videos created using AI.

Photo: Global Look Press/Marcus Brandt

65% of them depict crimes such as rape, torture and bestiality, compared to just 13% in 2024. Against this background, IWF experts insist on the need to update Internet security laws, as well as introduce strict rules for relevant companies.

Why is child porn increasingly being created with the help of AI

The sharp increase in the generation of prohibited content, especially the huge leap in the video segment, is directly related to the unprecedented availability, cheapening and rapid development of open neural network models, says Alexey Mironov, a leading specialist in the research development department of Stakhanovets, an expert in the field of information technology and information security, in an interview with Izvestia.

Naked untruth: in Russia they want to ban thousands of "undressing" AI services

What is the danger of spreading such resources and how are they being dealt with in other countries?

—If a couple of years ago serious computing power and deep programming knowledge were required to create a realistic video or deepfake, today attackers can use inexpensive cloud services and ready—made tools," the expert notes.

Photo: Global Look Press/Marcus Brandt

According to him, the emergence of powerful open video generation algorithms has reduced the technical barrier to entry for cybercriminals to almost zero. At the same time, the head of the Alliance for the Protection of Children in the Digital Environment, Elizaveta Belyakova, draws attention to the fact that the sharp increase in the number of materials with sexually abused children created using AI technologies is associated, among other things, with the effect of "scalability": one user can create thousands of images and videos in a short time. In addition, there is still a high demand for such illegal content in the shadow segments of the Internet, the expert says.

How creators of prohibited content circumvent the limitations of neural networks

In a situation with an increase in the volume of prohibited content generation, the spread of so-called custom models plays a separate role, explains Vladislav Puchkov, founder of the GEO Scout neural network monitoring platform. According to the expert, cybercriminals do not use public neural networks "as is", but rather train open source models or assemble pipelines (mechanics) from several tools.

"The development of AI for programming significantly accelerates this process — such systems allow even untrained users to bypass restrictions, implement modifications and automate the generation of prohibited content,— explains Izvestia's interlocutor.

Photo: IZVESTIA/Yulia Grigorieva

Protection circumvention, as Vladislav Puchkov points out, occurs in several ways: through local (offline) models without built-in moderation, through additional training on illegal datasets, as well as through so-called jailbreak approaches. In addition, there is a market for ready—made solutions, from private models to closed communities with instructions.

— In closed communities, which are actively used by intruders, both instructions and tools for circumventing defenses are distributed, — adds Elizaveta Belyakova.

What is the danger of the growth of child porn generation with the help of AI

The growth in the volume of prohibited content created using neural networks poses a serious threat, Elizaveta Belyakova emphasizes in an interview with Izvestia. Even if the materials are synthetic, they contribute to the normalization of violence against children, generate demand and can stimulate real crimes.

"In addition, technology allows you to create images that mimic specific children, which creates additional risks to their safety and reputation," the expert notes.

Photo: IZVESTIA/Pavel Volkov

Alexey Mironov also points out that AI allows attackers to create deepfakes using photos of real children taken from open social networks. This practice opens the way for violent cyberbullying, blackmail and extortion, which causes irreparable psychological trauma to victims and normalizes illegal behavior on the darknet.

The main danger, according to the expert, lies in scale: huge arrays of generated realistic content overload platform moderation systems and law enforcement agencies, making it critically difficult to find real victims of violence among terabytes of digital garbage.

How to deal with prohibited content from neural networks

In the coming years, we can expect a further increase in the volume of prohibited content generated using AI, Elizaveta Belyakova notes with regret. Technologies will evolve, become more realistic and accessible, and without systemic measures, this segment of illegal content may expand significantly.

Vladislav Puchkov agrees that the dynamics will worsen in the next few years.

— The video quality will continue to grow, and the cost of generation will decrease. The growth of the video format is particularly alarming, as it has a stronger effect on perception and is more difficult to detect automatically," he warns.

Photo: Global Look Press/Sebastian Gollnow

Effective fight against such threats, according to Alexey Mironov, requires the introduction of the principle of "AI against AI", when trained algorithms automatically detect and block illegal generation at the provider level, and all legitimate neural network content receives mandatory digital watermarks.

At the same time, from the information security side of the corporate sector, it is critically important to strictly control the IT infrastructure of companies and timely detect the launch of illegitimate AI models in the workplace, as well as to stop the generation or distribution of prohibited content, the expert notes.

Photo: Global Look Press/Fotostand/Gelhot

"Public and industry initiatives to combat child sexual exploitation, including online, are already developing in many countries," adds Vadim Glushchenko, Director of the Competence Center for Global IT Cooperation. — One of the most famous is INHOPE, an international network of hotlines that unites 52 countries, including Russia. She is engaged in the rapid identification and removal of illegal content related to children.

The key role is played by data exchange between Internet platforms and providers, as well as the development of common approaches to identifying prohibited content. At the same time, it is important that such interaction goes beyond individual countries — the involvement of international participants makes it possible to build a more coordinated and effective response, the Izvestia interlocutor concludes.