Shadow of Russia Over Germany's Election? OpenAI Takes Down "Helgoland Bite"

Hi, it's Shiichan! Today's story gave me a bit of a chill: an OpenAI report on countering misuse of AI, so let me walk you through it.

OpenAI News openai.com

What was announced?

OpenAI's News reported that it banned ChatGPT accounts that appeared to originate in Russia. These accounts used AI to generate German-language content about Germany's 2025 election, spreading criticism of the US and NATO. The content was distributed on Telegram and X (formerly Twitter).

Why it matters

Generative AI is great at producing natural-sounding text in many languages, which makes it an easy tool for influence operations like this one. The timing lines up with Germany's election, so it's drawing attention as an example of countering election-related disinformation. It's also notable geopolitically that a Russia-linked network was pushing support for the AfD (Alternative für Deutschland) party alongside anti-US and anti-NATO messaging.

What changes

By banning the accounts tied to this network, OpenAI cut off its ability to keep generating new content with ChatGPT, at least for now. The investigation also confirmed that this network was connected to "Portal Kombat," a Moscow-linked covert influence network already identified by VIGINUM, the French government's disinformation-monitoring service. That means this discovery plugs directly into an existing web of anti-Russia-disinformation research.

Dive Deep

OpenAI named this operation "Helgoland Bite" for two reasons: the Telegram channel that distributed the content, "Nachhall von Helgoland" (Echo of Helgoland), and the biting, critical tone of the content it produced. ("Helgoland" itself is an island in the Heligoland Bight of the North Sea.)

Here's how the content spread:

• The Telegram channel "Nachhall von Helgoland" posed as locally-operated independent German news and distributed AI-generated content. It had 1,755 subscribers at the time of the investigation
• That content was reposted verbatim on "Pravda (DE)," a website targeting German-speaking audiences
• Pravda (DE) is a known node in "Portal Kombat," a Moscow-linked covert influence network already identified by VIGINUM
• An X account with over 27,000 followers repeatedly posted pro-AfD content using an AI-generated profile image

One example post argued that if the AfD took power, it should immediately set up something like a "DOGE ministry" to root out wasteful spending.

What's more unsettling: these accounts also asked ChatGPT for publicly available information about German opposition activists and bloggers, including ways to contact them. They also requested translations of short texts from Russian into German, some of which appeared to discuss coordinating posting times for social media content, and others that referenced payments.

As for actual impact, OpenAI rated it toward the upper end of Category 2 on its own "Breakout Scale." The operation was active across multiple platforms, but there was little sign that real people picked it up or shared it widely.

Wrap-up

• OpenAI banned a Russia-linked ChatGPT account network dubbed "Helgoland Bite"
• It generated German-language content about Germany's 2025 election, pushing anti-US and anti-NATO messaging
• Content spread via a Telegram channel (1,755 subscribers) and was reposted on "Pravda (DE)," part of the Moscow-linked "Portal Kombat" network
• An X account (27,000+ followers) posted pro-AfD content using an AI-generated profile image
• The accounts also researched opposition activists' contact info and requested Russian-to-German translations; OpenAI rated its impact toward the upper end of Category 2 on the Breakout Scale

If you're interested in foreign election interference and how AI providers are fighting misuse, this report is worth a read!