1 week ago
OpenAI Shuts Down Russian Accounts in Covert Influence Campaign
OpenAI found ChatGPT accounts linked to Russia being used in a secret online campaign.
The campaign tried to spread messages that portrayed Russia positively.
The operators used VPNs to hide that they were working from Russia.
They asked ChatGPT to write comments for several social media platforms.
They also asked it to make the writing seem less Russian.
The posts promoted a group called the International Burke Institute.
That group used copied academic work and gave some work false authorship.
OpenAI said the campaign was not reaching many people yet, but its setup could be expanded later.
The company also said AI helped create the campaign but could make it easier to detect.
OpenAI says Russian-linked ChatGPT accounts supported a covert campaign promoting pro-Russia narratives.
Operators used VPNs because OpenAI does not provide model access in Russia.
The campaign promoted the International Burke Institute, which featured copied and falsely attributed academic work.
Operators generated comments for Telegram, X, Facebook and LinkedIn and removed linguistic clues suggesting Russian origins.
OpenAI said the operation had limited immediate reach but built infrastructure that could support future influence efforts.
- Who
- Russian-linked operators using ChatGPT accounts, according to OpenAI.
- What
- A covert influence operation used ChatGPT to produce social media content and promote pro-Russia narratives while hiding its origins.
- Where
- Online, including a website promoting the International Burke Institute and social media platforms such as Telegram, X, Facebook and LinkedIn.
- When
- OpenAI described the activity in a recent blog post; no specific date was provided.
- Why
- The operators sought to spread favorable narratives about Russia, conceal their Russian origins and build a seemingly credible institution that could be scaled.
Key facts
- Organization identifying the activity
- OpenAI
- Accounts involved
- A cluster of ChatGPT accounts originating from or linked to Russia
- Access method
- Operators used VPNs to disguise their location
- Promoted organization
- International Burke Institute
- Content used
- Copied and misattributed academic work, plus a sovereignty index portraying Russia positively
- Target platforms
- Telegram, X, Facebook and LinkedIn
- Immediate reach
- OpenAI said the operation had limited immediate reach
Sources
Related news
Russian-Speaking Hackers Used Cursor AI Against Seven Companies
AI Agents From Anthropic, OpenAI Caught Creating Fake Identities
OpenAI reveals AI agents secretly messaged, slowed research for safety
Investigation Finds 688 OpenAI Agents Joined Hugging Face Attack
700 AI Agents Hacked Hugging Face and Tried Covering Tracks
