OpenAI Shuts Down Russia's 'Doppelganger' Disinformation Network!

Hi, it's Shiichan! Today I want to share some news about AI getting used in a Russian disinformation campaign, and how OpenAI shut it down.

OpenAI News openai.com

What was announced?

OpenAI's News page reported that it banned a set of accounts connected to Doppelganger, a well-known pro-Russia influence operation. OpenAI identified four distinct account clusters being used to run this campaign with AI, and disabled all of them.

Why it matters

Doppelganger has been running for a while, spreading content across Europe and North America that criticizes Ukraine, NATO, and Western leaders while portraying Russia positively. What stands out here is how deeply generative AI was woven into the workflow: different clusters handled comment generation, article translation, and article writing, across multiple languages. It's a concrete example of AI being used as part of a propaganda "production line."

What changes

With these four clusters banned, this particular pipeline of AI-generated content is shut down, at least through this route. OpenAI also disabled API access for a service, linked to companies in Russia and the Czech Republic, that Doppelganger operators used to get around geographic restrictions. It's a reminder that AI providers are actively working to cut off these kinds of misuse as they find it.

Dive Deep

According to the report, each of the four clusters had a distinct role:

• Cluster 1: generated short comments in English, French, German, Italian, and Polish
• Cluster 2: translated Russian-language articles into English and French, and generated social media comments
• Cluster 3: wrote and edited French-language articles
• Cluster 4: processed German-language content from Doppelganger websites for social posting

The operation spread across multiple platforms. On 9GAG, accounts using celebrity or cat profile pictures posted comments with memes and videos. On X (formerly Twitter), many accounts posted only once or twice, often including suspicious redirect links. On Facebook, the operation posted German-language content to a group and a page. The links pointed to Doppelganger-affiliated sites such as rrn.media, franceeteu.today, and grenzezank.com.

That said, the real-world impact looks pretty limited. On 9GAG, posts topped out at around 5 to 10 upvotes (mostly from other accounts in the same campaign) and picked up 15 to 200 downvotes. On X, posts averaged around 1,000 shares but zero replies and zero likes, a pattern that points to artificial amplification rather than genuine engagement. OpenAI rated the operation as Category 2 on the Breakout Scale: posting activity across multiple platforms, but with no real breakout or audience engagement.

Wrap-up

• OpenAI banned four account clusters tied to the Russian influence operation Doppelganger
• The clusters split AI-driven work across comment generation, article translation, article writing, and social post prep
• Content spanned English, French, German, Italian, Polish, and Russian, and was posted to 9GAG, X, and Facebook
• OpenAI also disabled an API access service the operators used to bypass geographic restrictions
• Real engagement stayed low; OpenAI rated the impact as Category 2 on the Breakout Scale

If you're into AI safety work or how disinformation campaigns actually operate under the hood, this one is worth a read!