OpenAI Shuts Down Russian AI Abuse | Public Opinion Manipulation via Fake Think Tank

What you'll learn from this article

• How OpenAI discovered and blocked a Russian influence operation in August 2026
• How ChatGPT was exploited via VPN to spread disinformation
• A new manipulation tactic: using a fictitious think tank to fake credibility
• Why this small-scale operation could expand in the future
• The risk of similar tactics being used in Japan

What Was the Russian Influence Operation OpenAI Discovered?

On August 25, 2026, OpenAI (the American company that develops ChatGPT) announced that it had discovered an organized disinformation campaign originating from Russia and suspended the associated ChatGPT accounts.

In this influence operation, ChatGPT (an AI capable of writing text that reads like a human's) was exploited to mass-produce fake articles and social media posts designed to portray Russia favorably.

In other words, AI was used to make biased opinions — such as "Russia is a wonderful country" or "Western nations are in decline" — appear as though they had been written by genuine experts.

OpenAI identified and removed multiple accounts that had used VPNs (Virtual Private Networks: technology that disguises one's point of access) to circumvent access restrictions on connections from Russia.

The Sophisticated Tactics Used with ChatGPT

What deserves particular attention in this influence operation is how sophisticated the methods were. The operatives exploited ChatGPT in the following ways.

First, OpenAI restricts ChatGPT access from Russia. However, operatives used VPNs to disguise their country of origin and bypass these restrictions.

Next, after having ChatGPT write text in Russian, they gave instructions such as "Remove any Russian characteristics and make this natural English." This made it harder to detect that the text originated in Russia.

The resulting content was posted across multiple social media platforms, including Substack (a blogging platform), Telegram (a messaging app), X (formerly Twitter), Facebook, and LinkedIn.

By posting simultaneously across many platforms, they sought to maximize the number of people who would see the content.

Faking Credibility with a Fictitious Think Tank

At the center of this influence operation was a fictitious think tank (a policy research organization) called the "International Burke Institute" (IBI).

The IBI's website stated it was based in Israel, making it appear at first glance to be a legitimate research institution.

Even more strikingly, the names of prominent scholars Francis Fukuyama and Noam Chomsky were listed on the site without their permission.

According to OpenAI's investigation, of the 36 articles IBI published between September 2025 and May 2026, an astonishing 34 were plagiarized from other sites.

In other words, articles written by real scholars were copied without permission and presented as IBI's own research output.

IBI also published its own original metric called the "Sovereignty Index." This index was designed so that Russia would receive high scores, serving as a tool to portray Russia favorably.

Small in Scale, but Potentially "Groundwork" for Something Larger

The actual impact of this influence operation was limited. According to OpenAI, social media post view counts were low, and Telegram follower counts amounted to only around 10,000 to 20,000.

However, experts are wary that this small-scale operation may be "laying the groundwork" — meaning that even if the impact is minor now, it could develop into a large-scale disinformation campaign in the future.

Using AI tools, it is easy to fabricate a fake group of experts and make it appear to be a seemingly credible source of information.

Previous disinformation campaigns often relied heavily on large numbers of bot accounts (accounts that post automatically), which were relatively easy to identify.

However, tactics like the one used this time — impersonating a real research institution, publishing plagiarized academic articles, and using the names of prominent scholars without permission — are far more sophisticated and difficult to detect.

Impact on Japan and Future Countermeasures

While this particular incident originated in Russia, similar tactics could very easily occur in Japan as well.

For example, impersonating a fictitious Japanese research institute and mass-posting ChatGPT-generated articles in Japanese is something that can be done with relative technical ease.

Steps each of us can take as individuals include:

• Verifying whether a source is genuine by cross-checking across multiple sites
• When a prominent expert's name appears, checking whether that person is genuinely affiliated with the organization in question
• Being cautious of articles that present overwhelmingly one-sided views or use heavily emotional language
• Cross-referencing information with official sites and reliable news outlets

Companies like OpenAI are also advancing technological development to prevent AI abuse. However, technology alone cannot prevent it entirely.

Ultimately, the most important defense is for each of us — as recipients of information — to cultivate the habit of asking: "Can this information really be trusted?"

Summary

• In August 2026, OpenAI discovered and blocked an influence operation involving the abuse of ChatGPT, originating from Russia
• Access restrictions were bypassed via VPN to generate disinformation
• A fictitious think tank, "IBI," was created and the names of prominent scholars were used without permission
• 34 out of 36 articles were plagiarized, and a "Sovereignty Index" designed to portray Russia favorably was also published
• Currently small in scale, but potentially "groundwork" for a future large-scale operation
• Similar tactics could occur in Japan, making it important to develop the habit of verifying the authenticity of information

Generative AI is a convenient technology, but when abused it can become a powerful tool that leads us to make flawed judgments. This incident serves as a renewed reminder of the importance of information literacy in the age of AI.

---

This article is a cross-post from AI Friends.