OpenAI Busted Russia's 'Doppelganger' AI Op—It Was a Total Flop

OpenAI has detailed how it hunted down and disabled a sprawling Russian influence network, known as ‘Doppelganger,’ that was clumsily weaponizing its AI models. Reporting on the takedown, OpenAI revealed a campaign that was as ambitious in its targeting as it was pathetic in its execution. The operation spanned four distinct clusters of accounts, each acting like a dysfunctional department in a propaganda factory, churning out anti-Ukraine content in English, French, German, Italian, and Polish.

The first cluster focused on social media spam, posting on 9GAG and X. The tactics were straight out of a bot farm playbook—so much so that real users called them out. On 9GAG, accounts with cat and celebrity profile pictures would post a meme, only for three to five sock puppets to reply with “hahaha” or “lol”. These accounts were created on the same day and never interacted with anything else. The backlash was immediate; the posts were often downvoted into oblivion, attracting more critical comments than support. On X, the operation was even more transparent. Posts garnered roughly 1,000 shares but zero likes or replies—a statistical impossibility that screams inauthentic amplification. When a reply did trickle in, it was a user blasting the post as fake.

A second cluster acted as a translation bureau, converting Russian articles into English and French for the Doppelganger-linked site rrn.media, while a third farmed French content for franceeteu.today. A German-focused cluster ingested content from grenzezank.com to post in a Facebook Group. The operators tried to cover their tracks by accessing OpenAI’s API through a proxy service in Russia and the Czech Republic, a clear attempt to bypass regional bans. OpenAI has since disabled that service’s access. The content itself was a predictable mix of Kremlin talking points: portraying Ukraine as corrupt and on the verge of defeat, while painting Russia as an essential economic partner betrayed by out-of-touch Western leaders.

What’s genuinely striking here isn’t the sophistication of the operation, but its total failure to resonate. Using the Breakout Scale, OpenAI rated the impact a Category 2—content is posted across platforms, but authentic audiences couldn’t care less. The entire campaign appears to have generated near-zero positive organic engagement. This stands in stark contrast to the high-fear narratives often surrounding generative AI and election security. While the threat of AI-generated disinformation is real, the Doppelganger operation proves that generating the content is far easier than getting anyone to believe it. The social immune system of users calling out obvious bots worked as intended, making this a story less about AI’s destructive power and more about a costly, ineffective propaganda effort that simply bounced off its intended audience.