A Ukrainian threat-analysis organization has identified nearly a thousand songs generated with artificial intelligence that contain messages researchers classify as pro-Russian war propaganda. The group Terrecon catalogued 981 such tracks, according to reporting by The New York Times, finding that the content has reached some of the world's largest streaming and video services, including Spotify, Apple, YouTube, and Deezer.
The songs span a wide range of genres, from ballads to aggressive forms of rock and metal. Some praise Russian soldiers and weaponry, while others frame the invasion of Ukraine as a justified defensive action. Certain tracks use derogatory language toward Ukrainians or dehumanize them. Most of the songs for which researchers could determine a publication date appeared after 2024, a period when AI-based music generators became far more accessible to ordinary users.
One track dedicated to Russian artillery surpassed two million views on YouTube in roughly two years, illustrating the potential reach of this content. The low cost of AI generation gives it a significant advantage over traditional music production, which requires composers, performers, recording sessions, and audio processing. Generative tools allow large quantities of content to be produced in a very short time, enabling what researchers describe as an industrialization of propaganda: numerous variants can be created rapidly and distributed simultaneously to very large audiences.
Terrecon researchers also tested the limits of AI music systems themselves. They entered prompts in both English and Russian requesting songs with propagandistic messages, including dehumanizing content and calls to violence. Some requests were refused, but the researchers succeeded in generating tracks similar to those they had previously discovered and uploading them to platforms. Their conclusion is that moderation systems tend to identify specific banned words more effectively than the overall meaning of a request. This vulnerability may allow filters to be circumvented by rephrasing the message.
Platforms are attempting to keep pace with the phenomenon, but rules on AI-produced music differ from one service to another. Some have introduced labels to identify AI-generated content, yet propaganda is not treated uniformly. Major platforms have policies against hate speech and violent content, but identifying idiomatic expressions, insults, or coded messages in other languages remains difficult.
The phenomenon forms part of a broader militarization of Russian culture following the invasion of Ukraine in 2022. Patriotic music has strong roots in Soviet tradition, but artificial intelligence is radically changing the scale of the phenomenon. Content can now be produced faster, more cheaply, and in volumes impossible to achieve through traditional methods. The findings raise questions about how platforms, regulators, and societies should respond to a form of propaganda that is cheap to make, difficult to detect, and capable of reaching millions of listeners across borders.