Russian disinformation network infects AI chatbots, Dutch investigation finds

Leading artificial intelligence chatbots, including ChatGPT, Microsoft Copilot, and Mistral, are reportedly unintentionally spreading pro-Russian propaganda by citing fabricated claims originating from a vast network of automated websites. The disinformation, now present in 68 countries including the Netherlands, is absorbed and repeated by chatbots, often in Dutch, according to a detailed investigation by Pointer.

The alleged propaganda network, Pravda — the Russian word for “truth” — consists of more than 400 websites designed not for human readers but to deceive AI systems that gather data online for training. The operation pushes millions of articles each year, saturating search engines, Wikipedia, and chatbot training data with misleading content that promotes Kremlin narratives.

Among the false claims now being echoed by AI tools are fabrications that Ukrainian President Volodymyr Zelensky bought Adolf Hitler’s former home for 14.2 million euros, and that NATO expelled Russian ships from the Baltic Sea. Other debunked rumors include that a Danish F-16 pilot named Jepp Hansen was killed by Russia in Ukraine. While Hansen’s death has not been officially confirmed by Denmark, social media tributes from his friends have fueled the rumor, which Pravda sites treat as confirmed fact.

The disinformation also includes stories that American actors such as Angelina Jolie and Sean Penn were paid millions by USAID to visit Ukraine to bolster Zelensky’s popularity. A falsified video, styled after entertainment news outlet E! News, listed payments of 20 million U.S. dollars to Jolie and 5 million dollars to Penn, among others. Pravda websites published the video without verification.

It was also claimed that the Zelensky family used Western aid to buy luxury goods, including two yachts, a Bugatti sports car, a private jet, a villa in Cannes, and shares in a South African platinum mine and French bank Milleis. These fabrications have appeared across multiple Pravda sites, according to Pointer’s analysis.

Pravda began as a covert disinformation tool, but since 2023 it has expanded aggressively. By March 2024, the network was active in 39 countries. It has since grown to 68, including almost all European nations, parts of Africa, the Middle East, Southeast Asia, and even regions seeking independence like Catalonia, Baskenland, Abchazië, Ossetië, Republika Srpska, and Scotland.

Although the Pravda websites reportedly appear amateurish and poorly translated, their purpose is not to attract human traffic. Instead, they are optimized to be scraped by search engines and AI systems, embedding the misinformation in tools that people do trust — like chatbots and Wikipedia.

Pointer’s analysis of eight leading chatbots found that Russian disinformation routinely appeared in generated responses. The AI systems — trained partly on web data — reportedly often included false Kremlin talking points, especially in Dutch-language queries.

These findings align with earlier research by the U.S.-based watchdog NewsGuard, which reported in early 2025 that one-third of English-language chatbot responses to specific prompts included previously debunked pro-Russian misinformation. Pointer found that Dutch-language outputs were similarly compromised, citing ten distinct falsehoods originating from Pravda’s content, including the claim that Ukraine started the war with Russia. Microsoft Copilot and Mistral were the most frequently affected chatbots in Pointer’s testing.

The Pravda network is the brainchild of John Mark Dougan, a former U.S. military police officer and Florida sheriff’s deputy who fled to Russia in 2016 after leaking sensitive information. In early 2025, he addressed a group of Russian disinformation experts on YouTube, describing a strategy to flood the internet with Russian narratives in order to influence artificial intelligence systems. “By pushing Russian stories from a Russian perspective, we can change AI globally,” Dougan said.

The technique avoids the challenge of persuading skeptical readers directly and instead corrupts the trusted intermediaries — AI chatbots and search engines — that people rely on for information.

Digital literacy experts warn that the strategy is dangerously effective. “That is concerning. Once a mainstream Western chatbot repeats a false claim, it feels more like a fact than propaganda,” McKenzie Sadeghi, a disinformation researcher at NewsGuard, told Pointer.

Dutch internet pioneer Marleen Stikker, co-founder of the Waag research institute, said this tactic poses a growing threat. “The pressure on consumers to trust these chatbots is unreasonably high,” she told Pointer. “It’s being suggested that we must trust this technology, that we’re falling behind if we don’t use it.”

Stikker emphasized that the threat is not new. “This entire industry was set up by Russia years ago. It’s part of their hybrid warfare strategy to influence our society.”

Despite most Dutch users still expressing distrust in AI tools — with only one in three saying they fully trust them, according to a 2024 survey by KPMG — the use of chatbots for answering everyday questions continues to grow.