OpenAI has dismantled two separate influence operations originating from Iran that leveraged its ChatGPT tool to generate divisive political content. The campaigns targeted audiences in the United States and several other countries, using deepfake news articles and social media posts to sow discord. The company disclosed its findings in a detailed threat intelligence report published this week.
How the campaigns worked
The first operation, which OpenAI traced to a group known as Storm-2035, created ChatGPT accounts to produce long-form articles and commentary on topics like the Gaza conflict, Israel’s role in the Olympics, and US presidential candidates. The generated content was then disseminated through fake news websites and social media accounts that impersonated real news outlets. Some posts reached tens of thousands of engagements before being detected.
The second campaign was smaller in scale but more focused. It used ChatGPT to generate comments and short posts that pushed divisive narratives about US political figures and policies. This operation targeted users on platforms like Instagram and X, formerly Twitter, with material designed to amplify existing societal tensions. OpenAI says it did not find evidence that either campaign achieved significant viral reach or altered public opinion in a measurable way.
OpenAI’s response and broader context
OpenAI quickly suspended the accounts involved and shared detailed threat indicators with industry partners and law enforcement. The company emphasized that its detection systems flagged the activity before it could cause major harm, but acknowledged the growing sophistication of adversaries. This incident is part of a wider pattern of foreign influence operations using generative AI to create convincing disinformation at scale.
The timing is critical: the US presidential election is months away. Intelligence agencies have repeatedly warned that AI tools could be used to generate deepfake audio, video, and text to mislead voters. While these campaigns appear amateurish in some ways, their existence shows that adversaries are actively testing the boundaries of what AI can do to interfere with democratic processes. OpenAI’s proactive disclosure is a step toward transparency, but it also underscores the ongoing arms race between AI safety teams and malicious actors.
Experts note that influence campaigns often evolve after detection, shifting tactics and platforms. The long-term impact of these operations may be less about immediate reach and more about normalizing AI-generated disinformation in the public sphere. OpenAI stated it will continue to refine its detection methods and collaborate with researchers to stay ahead of these threats.
For those concerned about the integrity of online information, this case serves as a reminder that vigilance is necessary. The company’s actions provide a model for how tech firms can respond, but the broader responsibility falls on platforms, governments, and users to verify sources and think critically about what they see. As the election approaches, the ability to identify synthetic content will be a crucial skill.
Staying informed about the latest threats in AI security is essential. To learn more about how companies are protecting users from disinformation, visit Mylistingo.
The Netherlands, for internationals
One email a week: the week's news for internationals in the Netherlands, new practical guides and what changed in the cities. No spam, unsubscribe any time.
