OpenAI has banned accounts linked to a Russian-origin influence operation, dubbed "Helgoland Bite," that utilized AI to create German-language political content. This content, focusing on topics such as Ukraine, NATO, and German domestic issues, including the German 2025 election, was distributed across platforms like Telegram and X.
The operation involved generating short articles and social media comments, as well as requests for publicly available information on German opposition activists and bloggers. The accounts also sought translations from Russian to German, with some messages discussing coordination of social media posting times and referencing payments. This case study was initially published in OpenAI's June 2025 report.
Key Points
- OpenAI banned accounts originating from Russia for generating German-language political content.
- The content focused on Ukraine, NATO, German domestic issues, and the German 2025 election.
- Distribution occurred on Telegram and X, including a channel named "Nachhall von Helgoland" with 1,755 subscribers.
- Content was reposted verbatim on a domain affiliated with the Pravda network, identified as part of the "Portal Kombat" influence network.
- An X account with over 27,000 followers also distributed AI-generated content supporting the Alternative für Deutschland (AfD) party.
- The operation used AI-generated profile images for social media accounts.
- Accounts requested publicly available information on German opposition activists and bloggers, including contact methods.
Context
According to OpenAI, the "Nachhall von Helgoland" Telegram channel presented itself as locally operated independent German news. The content from this channel was regularly reposted on a Pravda (DE) website, which the French government's VIGINUM service previously identified as a node in the Moscow-linked covert influence operations network "Portal Kombat." The X account associated with the operation frequently posted AI-generated content that supported the AfD party and used an AI-generated profile image.
Why It Matters
This disruption highlights the ongoing use of AI models by state-linked actors for influence operations, demonstrating how AI can be leveraged for content generation, persona creation, and reconnaissance. Builders and practitioners should note the specific tactics, such as generating political commentary and researching individuals, to better understand and mitigate potential misuse of AI technologies.
What To Do
- Note the specific platforms (Telegram, X) and content types (political commentary, short articles, social media comments) used in this operation.
- Observe how AI was used for generating profile images and for basic research on individuals.
- Review the June 2025 OpenAI report for further details on this and similar operations.
- Consider the implications of AI-generated content being reposted by known influence networks like Pravda (DE).