Russia used ChatGPT to produce multilingual propaganda and promote a fabricated thinktank presented as Israeli, in an operation aimed at European audiences.
OpenAI uncovered a small group of operators in Russia who used VPNs and ChatGPT accounts to generate propaganda messages, which later appeared in English, German and French on Telegram channels, according to reporting by Le Monde on 25 August 2026. The messages portrayed the West as decadent and Russia as a grand alternative.
The same accounts were used to create written and visual material for the International Burke Institute, or IBI, a supposed Israeli analytical centre. The operation, also reported by Briefly, shows how the Kremlin’s information campaign against Europe has moved beyond openly identifiable propaganda outlets and towards the construction of apparently independent sources.
A thinktank built to look credible
IBI presents itself as an Israeli institution and lists organisations and prominent figures among its partners and experts, including the United Nations, UNICEF, the former US secretary of state Mike Pompeo and the academic Francis Fukuyama. But its published material was copied from other sources, while the named partners and experts had no connection with the institute. UNICEF confirmed that it did not co-operate with IBI.
Several people connected with IBI live in Israel and are of Russian origin. They had previously been involved in Dor Moria, an association supporting Russian-speaking Israelis. That background helped give the project an appearance of local legitimacy while concealing its role in a Russian influence operation.
The choice of an Israeli identity was significant. A foreign analytical centre can appear more authoritative to European experts, journalists and politicians than a clearly Russian propaganda outlet. Presenting Kremlin narratives as research from an overseas institution is intended to make them easier to circulate in professional and political debate.
Artificial intelligence expands the operation
The operators used ChatGPT to generate content in several European languages and attempted to hide signs of their Russian origin through VPNs. The material was prepared for distribution across Telegram, X, Facebook, LinkedIn and Substack, allowing a single operation to address different national audiences at speed.
That combination of generative artificial intelligence, social platforms and a fabricated institution offers a more flexible model than simply repeating messages from an official Russian outlet. Text, images and purported analysis can be adapted to the language and political context of individual countries, while the supposed source gives the material a veneer of independence.
OpenAI assessed the operation’s direct reach as relatively limited. Its importance lies instead in the infrastructure assembled around it: a false expert organisation, a network of social-media accounts, multilingual content and AI tools that can be reused or expanded. The exposure of IBI therefore does not mean that Russia’s information operations have ended.
A long-standing tactic with new tools
The creation of a fake analytical institution is not new. Documents stolen from and published online in 2025 described a plan by the Agency of Social Design, one of the Kremlin’s main private contractors for disinformation operations, to establish an “NGO/institute/thinktank in Israel”. Such structures allow Russian operators to place propaganda under a false flag and present it as independent Western research.
The newer element is the scale and speed that AI can bring to the method. Pseudo-expert organisations, fabricated studies and invented rankings can be fed into European media, research circles and political discussions without an obvious Russian origin. That makes attribution more difficult and creates additional opportunities to influence public opinion and political decisions.
Governments and technology companies will need to look beyond overt propaganda and identify the hidden infrastructure behind it: fake thinktanks, networks of supposed experts, false claims of institutional partnership and co-ordinated use of generative AI. Effective detection will depend on closer co-operation between governments, platforms, researchers and news organisations. The immediate reach of the IBI campaign may have been modest, but its repeatable design gives the Kremlin a template for future operations across Europe.
How should European governments and technology companies divide responsibility for detecting fabricated expert institutions before they influence public debate?
