- Leaked files show John Mark Dougan used Claude and Qwen to support a large pro-Russian influence campaign.
- Dougan told Claude his team wrote about 43,000 articles each month, with costs near $23,000.
- Investigators say the plan aimed to sway search results, spread false stories and shape what AI systems find online.
A leaked archive from a pro-Russian propagandist’s laptop has shown how he used AI chatbots in his work. The files belong to John Mark Dougan, a former Florida deputy sheriff who now lives in Russia.
They show how he used Anthropic’s Claude and Alibaba’s Qwen to write political content for Western readers. The goal was to spread pro-Kremlin stories.
Dougan’s Leaked Files Reveal a Plan to Flood the Web
NewsGuard published its findings on November 11, 2026. Its team reviewed emails, private messages, documents and chatbot conversations from Dougan’s computer. Dougan confirmed to NewsGuard over Signal that the files were real. He also dismissed concerns about the exposure, saying he did not care what people thought of him.
The files show Dougan wanted to shape search engine results and large language models (the AI systems behind chatbots). His plan was to fill the internet with pro-Russian articles. One document, titled “AI Server,” describes a plan to post rewritten news reports and AI-made content on a regular basis. He believed this would lift his websites in search rankings. It would also make their content easier to see. AI systems may read those articles when they gather information online.
One chat with Claude shows how big the effort was. Dougan asked the chatbot to prepare a list of expenses. According to the chat, he said his team wrote about 43,000 articles each month and could not afford all of it. The request shows how costly the operation had become for him. The documents also put his monthly running costs near two million rubles, or about $23,000.
The 43,000 figure came from Dougan himself. The available reports do not independently confirm that his sites published that many articles every month.
Claude and Qwen Helped Write Political Content
Qwen is Alibaba’s chatbot, while Claude comes from Anthropic. The leak holds dozens of chats with both from April to August 2026, covering about five months of his work. In them, Dougan asked for help with websites, political messages and content built to appeal to Western readers.
On June 8, he asked Claude for name ideas for a European free-speech website. The site was meant for French speakers. Claude gave several suggestions. It also explained why some English phrases would sound unnatural to French audiences.
Dougan used Qwen to prepare political messages too. One request asked for a Telegram post claiming Democrats planned protests against Donald Trump to disrupt the US midterm elections. Qwen wrote a reply that backed this claim. It also accused the media of helping to promote the story.
Other chats focused on claims about American biological weapons labs in Ukraine. Dougan gave Qwen a document about the labs. He then asked for material that supported anti-Western arguments. NewsGuard noted that the claims about US biological weapons sites were false.
The records also show Dougan asking Qwen how to make AI writing harder to spot. The chatbot told him to vary his sentence lengths. It also advised him to avoid common AI writing habits.
Anthropic’s usage rules ban deceptive information campaigns and the creation of fake identities. Alibaba also bars users from spreading false information with Qwen. Neither company answered NewsGuard’s questions about the chats.
Leak Fits a Wider Russian Influence Network
Dougan’s work fits into a larger Russian influence network. Microsoft has tracked a campaign called Storm-1516. It uses fake news websites, staged videos and linked online accounts to spread misleading political stories.
Other reports have also examined alleged Russian-linked online activity. Following Ukraine’s strikes on Yandex data centres, claims emerged about Russian bots, adding another dimension to concerns about the country’s online influence networks. Ukraine’s strikes on Yandex data centres sparked claims about Russian bots.
A 2025 technical report from France’s VIGINUM agency also named Dougan as part of Storm-1516. The report tied him to the CopyCop network. The agency described dozens of influence operations aimed at Western audiences.
Reports say two other investigations also studied the leaked material. One came out on November 10, and the other followed on November 11. The second described documents, messages and technical records. These offered a look at how Dougan’s operation worked. They also showed its links to Russian intelligence figures.
The files reveal growing frustration inside Dougan’s camp too. He complained about internet limits in Russia, tiredness and poor pay. He also showed interest in moving to Belarus. Later, he told NewsGuard that he was still in Russia.
The leak does not prove that Dougan changed any existing AI model. It also does not show that he forced chatbots to repeat his claims. It does expose a plan built on mass-produced online content. That plan tried to shape the information AI systems may find.
The findings raise fresh worries about how propaganda campaigns could use AI tools and huge amounts of online content. Such campaigns could shape how people understand major political events. NewsGuard’s review shows his plan clearly, but it does not show how well the plan worked.