OpenAI disrupts Russian and Iranian AI-assisted false-front influence operations

OpenAI says it banned two influence operations, one originating in Russia and one in Iran, after finding that the actors used ChatGPT to support false-front media activity rather than only low-grade social posting. The company published the findings on October 8, saying the campaigns paired AI assistance with older tactics such as fake personas, front organizations, media pitches and misleading internal reporting.

What changed

The Russia-origin activity, which OpenAI nicknamed Dark Clark, targeted countries across Latin America. OpenAI said the operators used ChatGPT mainly to draft internal reports, create campaign material and manage reporting around a self-described research platform called the Social Research Center. The company assessed the Russian case at Category 5 on the Breakout Scale, the first operation at that level it says it has disrupted since it began this reporting line.

The Iran-origin activity, called Bogus Bylines by OpenAI, used seven journalist personas to pitch long-form articles to online outlets. OpenAI said it identified almost 100 articles published or syndicated under those bylines across roughly a dozen outlets. The same operation also generated batches of social media comments, but OpenAI said the comment activity showed limited visible engagement compared with the article-placement work.

Why it matters

The practical warning is that AI can make a familiar influence operation look more professional and move faster, without making it fully automated. Editors, researchers, NGOs and companies should watch for provenance signals around contributors, not just whether a paragraph reads like AI. A polished pitch from a plausible byline can still be part of a covert campaign.

OpenAI also drew a distinction between operators' claimed impact and evidence that could be checked in public sources. That matters for defenders because internal reports from influence actors can exaggerate results. CyberOGZ's read is that the more important risk is not mass comment spam, but AI-assisted front identities that can survive basic editorial scrutiny long enough to reach real audiences.

What to watch

OpenAI said it shared information with relevant authorities and industry partners. The next useful signal will be whether platforms and publishers turn this kind of disclosure into repeatable checks for contributor identity, syndication chains and unusual clusters of article pitches. For readers, the decision rule is simple: when a new expert voice appears around a conflict or election, source history and transparent affiliation matter as much as writing quality.

Sources

• openai.com

Cover photo by Ryutaro Tsukata on Pexels, used under the Pexels License.