Mariusz Żabiński, president of the Institute of Socio-Political Technologies "Kybernetes" and a member of the Consultative Council for Resistance to International Disinformation at Poland's Ministry of Foreign Affairs, said Russia was using Wikimedia Commons to disseminate propaganda because the online repository has become an important source of multimedia content used by AI systems.

Wikimedia Polska, however, said the prevalence of Russian materials in some search results stemmed primarily from the availability of those files under free Creative Commons licences rather than documented, coordinated interference by Russia.

Wikimedia Commons is a freely licensed repository of images, audio and video files that operates on the same collaborative model as Wikipedia, allowing users to upload and edit content. The platform hosts more than 140 million multimedia files.

Żabiński pointed to search results for "annexation of Crimea" on the English-language version of Wikimedia Commons as an example. According to the expert, more than 40 of the 51 search results originate from the official website of the President of the Russian Federation or from Sevastopol Independent Television, a Russian broadcaster.

The results include recordings of rallies supporting Russia's 2014 annexation of Crimea and footage of Denys Berezovsky, the former commander of the Ukrainian Navy who pledged allegiance to the pro-Russian authorities in Crimea after the peninsula's occupation.

He also said image search results include photographs from the signing of the treaty incorporating Crimea into the Russian Federation, as well as slogans used by Russian President Vladimir Putin to justify the annexation.

According to Żabiński, the significance of such content extends beyond public access because Wikimedia Commons has become an important source of multimedia data for artificial intelligence.

"Wikipedia and Wikimedia Commons are among the most important sources for training and querying AI about knowledge about the real world," he said.

The Wikimedia Foundation has reported that downloads of multimedia content from Wikimedia Commons have increased substantially since 2024. According to the foundation, bandwidth used to download media files from the repository has risen by 50%, driven primarily by automated programs retrieving images for AI applications rather than by readers or media organisations.

Żabiński said advances in artificial intelligence have expanded the importance of multimedia repositories because modern language models increasingly process images, audio and video in addition to text.

"Support for the image layer was one of the most important steps in the evolution of language models and currently most of them are multimodal, i.e. they understand, or rather decode, what is contained in images, and based on this draws conclusions and generates further content," he said.

He also argued that metadata accompanying uploaded files could influence how both human users and AI systems interpret multimedia content.

"Much more interesting meta-information is a short description of what is in the photo. And this is a very large room for abuse, because it may influence how the recipient or the language model should interpret the file," he said.

Żabiński further warned that images could potentially contain hidden information embedded using steganography, although he did not present evidence that such techniques are currently being used on Wikimedia Commons.

"These may be instructions specifically ordering a change in the way of analysing the action or intended to trigger a specific action on the part of the language model," he said.

The Wikimedia Foundation did not respond to a request for comment from the Polish Press Agency (PAP) before publication.

In a statement to PAP, Wikimedia Polska rejected suggestions that the available examples demonstrated a coordinated Russian influence campaign.

"Based on the received examples and consultations with active Wikimedia Commons volunteers, we see no grounds to talk about documented, coordinated Russian interference in Wikimedia Commons. The increased presence of materials from Russian sources in the case of entries such as 'annexation of Crimea' results primarily from the licensing availability of these materials," the association said.

The organisation explained that some photographs and other materials published on the official website of the President of the Russian Federation are released under the Creative Commons Attribution 4.0 International licence. Because Wikimedia Commons accepts freely licensed material with educational value, such files may be uploaded to the repository provided they comply with the platform's policies. The association added that many other photographs from the same source are not available under open licences.

A Wikimedia Commons volunteer identified by the username "Cybularny", quoted in the association's response, said the repository accepts eligible material regardless of its source.

"On Commons we upload external materials that are made available under an acceptable free license, and since the website of the President of the Russian Federation provides free materials, we use them," the volunteer said.

Wikimedia Polska added that storing files originating from a particular institution does not imply endorsement of that institution's views or political narrative. The repository's purpose, it said, is to preserve freely reusable material that has educational value.

Żabiński argued that the prominence of Russian-source material could nevertheless shape perceptions of the annexation of Crimea by placing greater emphasis on geopolitical narratives than on the human consequences of the conflict.

According to the expert, if information about the suffering, losses and victims on the Ukrainian side becomes less visible in searches while Russian political messaging becomes more prominent, the humanitarian dimension of the conflict risks being relativised.

He also said awareness of the issue remained limited among researchers and the wider public, reducing efforts to address it.

"Meanwhile, the scale of activity of Russian entities in the Wikimedia repository will only intensify, reaching overwhelming proportions," Żabiński said. (PAP)

wm/ pś/ mhr/

tr. RL