OpenAI said it has identified and disrupted a set of influence operations tied to Russian and Iranian actors that used ChatGPT to support deceptive online campaigns, including the creation of fake articles, social media posts and fabricated personas. The findings add to mounting evidence that generative AI is becoming a practical tool for state-linked information warfare, even when the underlying operations remain relatively unsophisticated.
The company said the activity involved attempts to produce political content, commentary and outreach material designed to appear authentic while concealing the operators behind it. In some cases, the material was aimed at audiences in the United States and other countries, reflecting a broader pattern in which foreign actors use AI to scale influence efforts that once required larger teams of writers, translators and social media managers.
AI as a force multiplier
OpenAI's disclosure underscores a central shift in modern disinformation campaigns: the technology does not need to invent a new strategy to be dangerous. Instead, it can accelerate old tactics. A single operator can now draft dozens of posts, tailor messaging to different audiences, and rapidly rework language after content is removed or challenged. That makes influence operations cheaper, faster and more resilient.
The company said the Russian-linked activity included efforts that used ChatGPT to help craft content and personas intended to deceive schools, a politician and media organizations, according to reporting cited by NBC News. Iranian-linked actors, meanwhile, were said to have used the tool in attempts to plant false stories in multiple U.S. media outlets, according to CNN. OpenAI said it disrupted the accounts and associated activity after detecting policy violations and signs of coordinated abuse.
The cases are notable not because they represent a technological breakthrough by the operators, but because they show how quickly adversaries are adapting to mainstream AI systems. The campaigns appear to have relied on the model for drafting and editing rather than for autonomous decision-making. Even so, the use of a widely available chatbot in covert operations raises difficult questions about how platforms should monitor abuse without restricting legitimate use.
A familiar threat, faster
Foreign influence operations are not new. Russia, Iran and other states have long used online propaganda, fake personas and media manipulation to shape public debate, sow confusion and amplify political divisions. What has changed is the industrial efficiency of the process. Generative AI can produce fluent text at scale, imitate tone and style, and generate content in multiple languages with minimal effort.
That matters because the success of influence operations often depends less on technical sophistication than on volume, persistence and timing. AI can help operators flood platforms with content, test which narratives gain traction, and quickly pivot when a story is exposed. It can also make low-budget operations look more credible by polishing grammar, localizing language and producing plausible but false supporting material.
OpenAI's findings also highlight the growing challenge for news organizations and public institutions. Fake stories can be drafted to resemble legitimate reporting, while fabricated outreach can be tailored to specific targets. Schools, local officials and smaller media outlets may be especially vulnerable because they often lack the resources to verify every unsolicited message or suspicious submission.
Detection and deterrence
The company said it has taken steps to remove the accounts and block the activity, part of a broader effort by major AI firms to identify abuse and publish threat intelligence about malicious use of their systems. That transparency is increasingly important as governments and researchers seek to understand how generative AI is being weaponized.
Still, disruption alone is unlikely to solve the problem. As long as AI tools remain broadly accessible, state-backed and criminal actors will continue probing for ways to use them in influence campaigns, phishing, fraud and other forms of deception. The practical response will require a mix of platform enforcement, media literacy, stronger authentication practices and closer coordination between technology companies, governments and newsrooms.
For policymakers, the OpenAI report is a reminder that the debate over AI safety is no longer confined to hypothetical future risks. The technology is already being used in live information operations, and the scale of the threat is likely to grow as models become more capable and easier to deploy. The question is no longer whether foreign actors will use AI in influence campaigns, but how quickly defenders can adapt to keep pace.