OpenAI Bans Russian and Iranian False Front AI Ops

Overview

OpenAI has banned two distinct influence operations—one originating from Russia and another from Iran—that utilized its artificial intelligence models to construct sophisticated "false front" entities. According to a safety report published on October 8, 2026, these actors employed a combination of AI-generated content and traditional operational techniques to launder geopolitical messaging into target audiences.

The Iranian operation maintained a roster of seven fabricated journalist personas, which it used to pitch long-form articles to small and medium-sized online outlets globally. In a parallel case, Russian actors appear to have co-opted unwitting individuals in Latin America to manage an in-ground "think tank," using AI to draft internal reports and create misleading documentation.

"Both operations were multi-dimensional, trying a range of different tactics to achieve their goals," OpenAI stated. "As well as long-form articles, the Iranian operation generated batches of social media comments... both operations also made heavy use of AI to draft internal reports."

This action marks a significant escalation in the company’s efforts to disrupt covert influence campaigns. For the first time since OpenAI began systematically reporting on such threats, it identified and disrupted a Category 5 operation on its IO Breakout Scale, indicating a high potential for widespread reach and impact.

Scale and Sophistication of Threats

The two banned operations demonstrated a level of complexity that closely mirrored pre-AI era influence campaigns, but with significantly enhanced efficiency and scale. The Iranian group not only produced long-form journalism but also generated coordinated social media commentary, primarily focused on topics related to the US-Iran conflict. The operators used questionable or outright deceitful methodologies in their internal reporting to exaggerate the effectiveness of their campaigns.

The Russian operation, similarly, went beyond simple content generation. It created fake "leaked" documents and audio scripts, some of which were identified spreading across the internet. OpenAI noted that the Russian actors relied heavily on AI for drafting internal reports, using the technology to maintain the illusion of a legitimate research organization.

"What is most striking about these operations is that they closely resembled complex influence operations of the pre-AI age, but used AI to make some of the workflows easier."

OpenAI drew parallels between these modern AI-assisted operations and historical fronts. The Iranian journalist personas bore a resemblance to "Alice Donovan," a fictional identity linked to Russian military intelligence that published articles in Western outlets between 2016 and 2017. Similarly, the Russian operation shared characteristics with "PeaceData," a fake news outlet from 2020 that recruited unwitting journalists worldwide.

Impact and Reach Assessment

OpenAI’s IO Breakout Scale rates influence operations on a scale from 1 to 6, based on their potential to disrupt information ecosystems. The Russian-origin operation was assessed as a Category 5, the highest tier of disruption open to the company’s current assessment framework. This is the first Category 5 operation the company has disrupted since it began its public reporting initiative.

The Iranian operation was rated a Category 4. Both operations succeeded in placing their content into mainstream media outlets, rather than relying solely on decentralized social media distribution. This strategy is consistent with OpenAI’s broader findings: operations that infiltrate real media outlets tend to have the highest potential reach and impact compared to those that rely exclusively on fake social media accounts.

"Operations which used more than one distribution method... are classified by the method which appeared to constitute the core of the operation."

The ability to land content in legitimate news outlets represents a critical vulnerability that threat actors are increasingly exploiting. By using AI to improve linguistic fluency, editorial quality, and operational speed, these actors can deceive unsuspecting editors and employees, planting content before audiences who remain unaware of the true motivations behind it.

Operational Breakdown: The Russian Campaign

The Russian cluster involved a group of ChatGPT accounts originating from Russia, despite OpenAI’s existing restrictions on accessing its models from that region. The primary objective of this campaign appeared to be undermining Ukraine’s reputation across Latin America, with secondary goals influencing local political outcomes in countries such as Argentina and Bolivia.

Most operators prompted in Russian, though one used Spanish while remaining located in Russia. The campaign centered on a self-described "research platform" known as the Social Research Center (SRC). The actors controlled this entity through a fake persona named "Mia Clark." OpenAI has nicknamed the operation "Dark Clark" in reference to this alias.

The core activities of Dark Clark included:

• Writing internal reports on the operators’ own activities and attributing them to other actors to inflate perceived success.
• Creating propaganda content tailored for Latin American audiences.
• Drafting performance reports for the SRC to maintain the facade of a legitimate research institution.

Operational Breakdown: The Iranian Campaign

The Iranian operation distinguished itself by maintaining a stable of seven distinct journalist personas. These identities were used to pitch lengthy, professionally written articles to small and medium online outlets around the world. The goal was to embed state-aligned narratives into independent media platforms, thereby granting them an air of credibility.

In addition to article generation, the Iranian group produced coordinated social media comments to amplify specific talking points. Like their Russian counterparts, they used AI to streamline their workflow and present a more polished, professional appearance to their targets.

The Strategy of Disclosure

OpenAI emphasized that while these false-front operations are sophisticated, their covert nature makes them vulnerable to exposure. History has shown that when such operations are publicly identified, they often cease activity. The previous fronts of "Alice Donovan" and "PeaceData" both ended after being exposed.

"That’s why our goal in reporting these false-front operations is to make further research and disruption easier—and make continuing the operations harder."

By publishing detailed analyses of these campaigns, OpenAI aims to help regulators, industry peers, and society at large understand the specific vulnerabilities threat actors exploit. As AI tools become more accessible, the barrier to entry for conducting such operations lowers, making robust defense mechanisms and public awareness essential for maintaining the integrity of information ecosystems.