OpenAI disrupts Russian and Iranian AI influence operations

OpenAI banned Russian and Iranian covert campaigns that used ChatGPT to run fake journalists and a Latin American think tank.

OpenAI said it banned two covert influence operations, one originating in Russia and one in Iran, that used ChatGPT and other AI tools to support false-front media and think-tank campaigns. The Russian cluster, nicknamed Dark Clark, used VPNs and a persona named Mia Clark to run the Social Research Center, focused on harming Ukraine’s image in Latin America, and spread fake audio tied to Ecuador and Bolivia that prompted official denials. OpenAI rated it Category 5 of 6 and called it both the highest-rated operation it says it has disrupted and the most complex front identity it has disrupted in two and a half years, without naming a Russian agency. A separate Iranian-linked effort used seven journalist personas—CyberScoop says at least seven—to pitch long-form articles and social-media comments about the U.S.-Iran war, placing nearly 100 articles across about a dozen outlets, and was rated Category 4 of 6. Both operations placed content in mainstream outlets and used AI heavily to draft articles, fake leaks, and internal reports that inflated results; neither was attributed to a named intelligence service.

• OpenAI banned Russia-origin and Iran-origin covert influence accounts that used ChatGPT and other AI tools.
• The Russian cluster, nicknamed Dark Clark, used VPNs and a persona named Mia Clark to run the Social Research Center, a Latin America think tank aimed at harming Ukraine’s image.
• OpenAI rated Dark Clark Breakout Scale Category 5 of 6, calling it the most complex front identity it has disrupted in two and a half years and the highest it says it has disrupted.
• Dark Clark spread fake audio tied to Ecuador and Bolivia that drew official government denials.
• The Iranian effort used seven, or at least seven, fake journalist personas to pitch articles and social-media comments about the U.S.-Iran war, placing nearly 100 articles across about a dozen outlets, and was rated Category 4 of 6.
• Both operations placed content in mainstream outlets and used AI to draft articles, fake leaks, and internal reports that exaggerated effectiveness.
• Neither cluster was attributed to a named intelligence service.