Internal documents leaked from the Russian private agency Social Design Agency (SDA) reveal a plan known as "Project 2026" to manipulate the information base that search engines and AI chatbots rely on. Bloomberg News verified the obtained files and reported on the plan on June 23, 2026.
June 23, 2026 · Leaked Internal Documents
"Project 2026": Russia's Plan to Poison the Data Behind AI Chatbots and Search
Leaked files attributed to the Social Design Agency describe building Wikipedia-style clones, fake think tanks and self-filling knowledge bases — engineering propaganda upstream , into the reference sources that feed search engines and large language models.
73
leaked internal documents
May '23– Apr '26
period the files were created
32
domains seized in 2024 "Doppelganger" case
0
large-scale deployments confirmed publicly
The shift: from social feeds to the reference layer
The plan moves disinformation upstream — into the sources that AI and search treat as facts.
Fake sourcesclones · think tanks · self-filling bases
→
Search index &
→
Chatbot answerspropaganda surfaces "inconspicuously"
Assets described in the documents
A Wikipedia-style clone aimed at Germany
A "World Center for Strategic Studies " in English, German, French & Spanish — domain registered March 30, 2026, publishing anonymous analysis
AI-driven self-filling knowledge bases ; servers and web shells reportedly already prepared for a German-targeted baseCoordination with Russia's presidential administration, using the term "cognitive strikes "
What's unconfirmed
These are planning documents . No large-scale real-world deployment or user impact has been made public. A 2025 study found chatbots citing Russian state media for some queries — but no direct link to this operation is confirmed.
What's verified
European researchers and authorities reportedly verified the files' authenticity by content and format. SDA — already under U.S., U.K. and EU sanctions — did not comment.
The method targets "data voids" — gaps where reliable information is scarce — to seep falsehoods into AI answers via sites masquerading as encyclopedias and think tanks.
How AI providers and search platforms verify the reliability of their reference sources is set to become a defining challenge.
Continue reading The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.
Already purchased? Sign in ✓ Signed in — this article isn’t included in your current plan.Unlocking the full article…