OpenAI Disrupts Sophisticated AI-Driven Influence Campaigns from Russia and Iran

In an era where digital disinformation has become a primary weapon of statecraft, OpenAI has revealed a significant victory in the battle against AI-powered propaganda. The artificial intelligence giant announced on Thursday that it had successfully identified and dismantled two covert influence operations orchestrated by state actors from Russia and Iran. These campaigns were designed to weaponize ChatGPT to create a veneer of legitimacy, allowing foreign agents to masquerade as independent journalists and academic researchers to sway global public opinion on volatile geopolitical conflicts.

The Iranian operation was characterized by a high degree of persona-building. According to the report, the campaign deployed seven distinct fake identities, all posing as freelance journalists from Western nations. By fabricating professional backgrounds and credentials, these accounts were able to infiltrate several small-to-medium-sized digital media outlets specializing in international relations and Middle Eastern affairs. These outlets, which often lack the rigorous vetting processes of major news organizations, became conduits for nearly a hundred long-form articles designed to push specific pro-Iranian narratives. The activity peaked following the escalation of tensions between the U.S. and Iran in March, signaling a strategic alignment between the propaganda efforts and real-world political crises.

One specific persona, identified as "Evan B. Hoskins," claimed to be a U.S.-based freelance writer with a presence on X (formerly Twitter) and Instagram. However, OpenAI's forensic analysis uncovered a starkly different reality. Metadata and connection logs revealed that the X account was accessed via a West Asian Android application, while the Instagram account was traced directly to Iran. These digital footprints exposed the deception, leading to the immediate suspension of the accounts.

Simultaneously, OpenAI uncovered a separate, equally insidious operation directed by Russia. Unlike the Iranian campaign, which relied on fake individuals, the Russian effort focused on institutional fraud. The actors established a fraudulent think tank titled the "Center for Social Research" based in Latin America. To maintain a facade of local authenticity, the operation recruited unsuspecting local employees in the region to produce content. The primary objective was to damage the international reputation of Ukraine within Latin American political circles and interfere with local political dynamics by spreading tailored disinformation.

OpenAI described these events as the most sophisticated covert identity operations it has encountered in the last two and a half years. The complexity lies in the shift from simple bot-driven spam to the creation of holistic, believable ecosystems—including fake careers, institutional affiliations, and strategic placement in niche media. This evolution suggests that state actors are no longer just using AI to generate text, but are integrating it into broader psychological operations to bypass traditional disinformation filters.

This incident serves as a stark reminder of the dual-use nature of large language models. While these tools provide immense value for creativity and productivity, they can be repurposed by national governments to industrialize the production of propaganda. By automating the creation of nuanced, high-quality articles that mimic the style of professional journalism, bad actors can flood the information space with convincing falsehoods at a scale previously impossible for human operators alone. OpenAI has since banned all involved accounts and continues to refine its detection mechanisms to prevent the misuse of its technology for geopolitical manipulation.