Leak reveals how former US marine tried to trick Claude into spreading pro-Russian propaganda
John Mark Dougan, a.k.a. BadVolf, has also targeted Alibaba’s Qwen.
Image by NurPhoto via Getty Images
Follow us
- Leaked files show John Mark Dougan used Claude to generate pro-Russian articles aimed at American readers.
- A prompt said the operation produced 43,000 articles a month and cost about $23,000 monthly.
- Dougan also used Alibaba's Qwen and asked how to avoid AI-detection tools.
Key Takeaways by nexos.ai, reviewed by Cybernews staff.
Leaked documents revealed how John Mark Dougan, a former US marine turned Russian propagandist, tricked AI into pushing pro-Russian narratives to American readers.
The files obtained from Dougan’s laptop showed how he used AI to generate tens of thousands of articles per month praising Russia, so they could be accessed by other AI chatbots.
For this, Dougan used Anthropic’s Claude to generate articles that would seem legit to American readers.
By using AI tools to create websites and content that spread pro-Russian narratives, Dougan violated the chatbot’s terms of service.
The methods of Dougan’s work were revealed after his emails, Telegram and Signal messages, and AI prompts were leaked and shared on a website imitating his own. The link to the website was shared via his hacked X account.
Dougan confirmed to NewsGuard that the leaked content was legit, adding:
“Everyone knows I was hacked. Can happen to anyone. Believe me nobody gives a shit. “Publish my laptop. I don’t care… I really don’t care what people think of me.”
Infecting AI chatbots
One of the documents revealed how he used AI to manipulate search algorithms and LLMs by constantly updating news articles that favor Russia with AI-generated content used to increase the articles' ranking and readability.
For this to work, the former US police officer would generate tens of thousands of articles every month, with the “operation” costing around 2 million rubles, or $23,000 per month.
“Currently we are writing 43k articles a month and we can’t possibly afford all of that,” Dougan said in a prompt asking Claude to create the list of expenses.
The leaked information showed which AI chatbots the man used to interact with. These include Anthropic’s Claude and Alibaba’s Qwen. He also used Open WebUI, a platform allowing users to interact with AI models on “their own terms.”
One example shows how Dougan used Claude to push his narratives, thereby violating Anthropic’s policies.
Dougan asked Claude to create a “free speech site for Europe,” targeting French speakers.
"I like speakup but I don't know if they would understand it in French,” shared Dougan, with Claude explaining that the suggested name wouldn’t work with French audiences.
According to Anthropic’s policy, the company doesn’t allow its AI model to be used to create misleading information, create fake people or accounts, or falsely attribute content.
Despite Alibaba’s policy against using Qwen to spread misinformation, Dougan turned to the AI, asking how he could write text that wouldn’t be detected by AI detectors.
Stay updated with our latest stories and follow us on social media
Be the first to discover new stories, ideas, and updates from our team.
Dougan’s career as a Russian propagandist
Dougan is a former US marine who served as a deputy sheriff in Palm Beach County, Florida, and later in Windham, Maine, but was dismissed due to misconduct and harassment.
In 2016, he was under FBI investigation for leaking confidential police records.
He then fled to Russia, where he received political asylum after being indicted on 21 felony charges in Florida. In 2023, Dougan was granted Russian citizenship.
It’s been reported that Dougan had ties with Russia's military intelligence service (the GRU), which funded his propaganda services.
Dougan, who also goes by the pseudonym “BadVolf,” created and managed fake news outlets disguised as legitimate American news sites to spread disinformation and false narratives directed towards the US.
