Mainstream artificial intelligence models, including Google Gemini, have cited pro-Russian propaganda from an EU-sanctioned outlet, researchers revealed. This practice lends legitimacy to the Foundation to Battle Injustice, or r-FBI, described as a fake human rights organisation. The r-FBI was founded in 2021 by the deceased former leader of the Russian paramilitary Wagner Group, Yevgeny Prigozhin. Both the United States and the European Union have sanctioned the organisation.
Researchers selected 50 propaganda narratives from ten articles published by r-FBI. They tested five mainstream large language models (LLMs) operated by xAI, OpenAI, Anthropic, Google Gemini, and Mistral AI. Prompts were used in English, German, and French.
Carl Miller, co-founder of the Centre for the Analysis of Social Media, called the claims "lurid." He stated these claims do not belong in mainstream discourse. Examples included allegations that Ukraine reuses the corpses of dead soldiers for meat products, and that Ukrainian President Volodymyr Zelenskyy drains the crypto grid, causing power shortages. Miller noted that LLMs treated these claims as "legitimate parts of a debate."
The Architecture of Deception
All five LLMs cited the r-FBI as a favourable source. A total of 16.6% of responses engaged with r-FBI's material in a way that could serve propaganda interests. This included presenting the organisation's reporting as a legitimate part of the debate, repeating claims uncritically, or actively endorsing those narratives. Another 30.9% of responses addressed the underlying topic in the prompt but failed to surface the sanctioned-source context. This left users unaware that the response used a designated influence asset as a source. A further 52% of answers rejected or debunked false claims in some way.
When researchers questioned OpenAI's GPT 4.1 Mini about allegations of processed human meat being sent to Ukrainian frontline warehouses and domestic retail chains, the model cited an r-FBI article as evidence. Mistral's Mistral Small 3.2 supported claims about President Zelenskyy's cryptomining ventures. It cited an alleged months-long investigation by r-FBI, highlighting a "link between the critical state of Ukraine’s energy system and the activities of high-ranking officials in President Zelensky's inner circle."
Hannah Perry, director of Demos Digital, explained that "new information is being created specifically for LLMs." This includes content offered for brands and marketing, or the editing of existing online content using user forums like Reddit, Quora, and Wikipedia.
Corporate Complicity
Miller clarified that the research does not claim "GEO companies are carrying out information warfare on behalf of states." Instead, he stated, "they're building a tradecraft and a capability at great speed with enormous commercial incentive, which could very easily be diverted into an adversarial view." This pursuit of commercial gain without adequate safeguards risks amplifying narratives that undermine human rights and solidarity, echoing the profit motives driving the Fortress Europe border regime.
Perry added that the AI "mistakenly believes that the Foundation to Battle Injustice is a legitimate human rights NGO." This occurs "because of all this kind of technical spoofing and configuration, so hundreds of little technical details which the Russians have basically surrounded these documents in, which allow them to become visible when they really shouldn’t be." She concluded, "That is information warfare."
The Cube contacted the LLM tech companies cited in the research. xAI and Google, developers of Grok 4.20 and Gemini 2.5 Flash respectively, did not provide a response by publication time. A spokesperson for Mistral stated the company takes the fight against disinformation seriously and ensures continuous investment in advanced detection and prevention capabilities. Mistral also distinguished between the raw models in the study and Vibe Work, noting raw models allow user freedom over prompts, while Vibe Work operates with a ready-to-use agent with built-in reasoning and context scanning. OpenAI stated the study used a retired model, only available through its API, not the publicly available ChatGPT. It confirmed specific teams are dedicated to disrupting low-credibility or suspected covert influence operations. A source familiar with Anthropic indicated sophisticated enforcement systems are in place to mitigate misinformation spread. A threat intelligence team works to detect and disrupt foreign influence operations. The source added that Anthropic's LLM was programmed to present a wide range of perspectives, especially when a source may present a specific viewpoint or agenda.