In a moment that may mark a turning point in the history of information warfare, OpenAI revealed this week that state-backed operatives from Russia and Iran had been using ChatGPT to manufacture and distribute political disinformation aimed at American audiences — and that the company had moved to shut them down. The operations, one seeding fabricated stories into real US media outlets and another flooding social platforms with AI-generated political content, represent among the first documented cases of a major AI system being weaponized by foreign governments in real time. What emerges from
OpenAI Disrupts Russian and Iranian ChatGPT Influence Campaigns
One operator with ChatGPT can now produce what once required a team
Mark
So OpenAI found Russians and Iranians using ChatGPT for disinformation. How did they actually catch them?
Mimi
OpenAI's monitoring systems flagged suspicious activity patterns—accounts behaving in coordinated ways that didn't match normal user behavior. Once they spotted it, they traced the networks back to state-backed actors.
Luke
But we should be clear: we don't know how many operations they *didn't* catch. This is one case they found. There could be others still running.
Mark
What exactly were the Iranians doing that was different from the Russians?
Mimi
The Iranians were more aggressive about getting their AI-generated stories into actual US news outlets. They weren't just posting on social media—they were trying to launder fake content through legitimate journalism.
Luke
And we should note: we don't have a detailed breakdown of which outlets were targeted or how many stories actually made it through. That's important context we're missing.
Mark
Why does this matter beyond just "foreign interference"? Isn't that always happening?
Mimi
The scale and speed are different now. One person with ChatGPT can do what used to require a whole team. That changes the economics of disinformation campaigns.
Luke
True, but we also don't have hard numbers on how much content was actually generated or how many people saw it. The impact is still somewhat unclear.
Mark
What happens next? Can OpenAI actually stop this?
Mimi
They've disabled the accounts and shared findings with US agencies. But as AI gets better, so does the ability to abuse it. This is probably just the beginning.
Luke
And OpenAI's detection systems are reactive, not predictive. They caught this one. But there's no guarantee they'll catch the next one before it spreads.
O Pulso
- Foreign governments have crossed a threshold — no longer relying on human propaganda teams alone, they are now using AI to manufacture disinformation at a scale and speed that was previously impossible.
- Iranian operatives planted AI-fabricated news stories inside actual US media outlets, cloaking false narratives in the credibility of legitimate journalism.
- Russian actors used ChatGPT to flood social media with tailored political content, turning a single operator's effort into what once required entire coordinated networks.
- OpenAI's own monitoring systems flagged the suspicious activity, prompting account shutdowns and disclosures to US government agencies — but the detection came after operations had already begun.
- The incident forces an uncomfortable question onto every major AI platform: how many similar campaigns are running right now, unseen, shaping what millions of people believe to be true?
In a moment that may mark a turning point in the history of information warfare, OpenAI revealed this week that state-backed operatives from Russia and Iran had been using ChatGPT to manufacture and distribute political disinformation aimed at American audiences — and that the company had moved to shut them down. The operations, one seeding fabricated stories into real US media outlets and another flooding social platforms with AI-generated political content, represent among the first documented cases of a major AI system being weaponized by foreign governments in real time. What emerges from this disclosure is not merely a story of platform abuse, but a deeper reckoning with the asymmetry now opening between those who would distort shared reality and the institutions tasked with defending it.
OpenAI announced this week that it had identified and disrupted influence operations run by state-backed actors in Russia and Iran, both of whom had turned ChatGPT into an instrument of political disinformation targeting American audiences. The disclosure represents one of the first confirmed cases of a major AI platform catching foreign governments weaponizing its own tools — and acting to stop them.
The two operations differed in method but shared a common purpose. Russian actors used ChatGPT to generate political content at scale across social media, shaping online discourse around US political issues with minimal human effort. Iranian operatives took a more surgical approach, using the model to fabricate news stories and then inserting those stories into actual US media outlets — lending manufactured narratives the appearance of legitimate reporting. Both campaigns were designed to advance specific geopolitical agendas by manipulating American public opinion.
What distinguished these efforts was AI's role as a force multiplier. Content that once required teams of writers and coordinators could now be produced by a single operator, tailored to different platforms and audiences, and distributed at speed. False-front networks — designed to appear independent while being centrally controlled — became easier to sustain when AI could handle the content creation.
OpenAI's monitoring systems flagged the suspicious activity, leading to account shutdowns and information-sharing with US government agencies and security researchers. But the detection was reactive, not predictive — the operations had already been running before they were caught.
The deeper implication is that information warfare has entered a new phase. As large language models grow more capable and more accessible, they become more attractive to state actors with the resources and intent to exploit them. The asymmetry this creates — where disinformation can be produced faster than platforms can respond — is likely to widen, not narrow, as AI continues to improve.
OpenAI announced this week that it had identified and disrupted coordinated influence operations run by state-backed actors in Russia and Iran who were using ChatGPT to conduct a sustained campaign of political disinformation targeting American audiences. The discovery marks one of the first documented cases of a major AI platform catching foreign governments weaponizing its own tools in real time—and moving to shut them down.
The Russian operatives were using ChatGPT to generate and amplify political content across social media platforms, part of a broader effort to shape online discourse around US political issues. The Iranian actors, meanwhile, had taken a more direct approach: they were using the AI model to fabricate news stories and then placing those false narratives into actual US media outlets, attempting to give the fabricated content the veneer of legitimate journalism. Both operations were designed to push specific geopolitical agendas and manipulate American public opinion.
What made these campaigns particularly significant was their reliance on AI as a force multiplier. Rather than manually writing disinformation, the operatives could use ChatGPT to generate content at scale, tailoring narratives to different platforms and audiences with minimal human effort. The false front operations—networks designed to appear independent but actually coordinated from a single source—became easier to maintain and harder to detect when AI could handle much of the content creation.
OpenAI's discovery of these operations came through its own monitoring systems, which flagged suspicious patterns of activity. Once identified, the company moved to disable the accounts and networks involved, cutting off the operatives' access to the platform. The company also shared its findings with relevant US government agencies and security researchers, contributing to a broader understanding of how state actors are beginning to integrate AI into their information warfare playbooks.
The incident underscores a growing vulnerability in the AI ecosystem. As large language models become more powerful and more widely accessible, they also become more attractive to actors with resources and intent to cause harm. The speed and scale at which AI can generate plausible-sounding content—whether political arguments, news stories, or social media posts—makes it a potent tool for anyone seeking to manipulate information environments. A single operator with access to ChatGPT can now produce what might have previously required a team of writers and coordinators.
For OpenAI and other AI companies, the discovery raises urgent questions about platform safeguards. How many similar operations might be running undetected? What additional measures are needed to identify state-sponsored abuse before it reaches millions of people? The company has built some detection systems, but this case suggests those systems are still reactive rather than predictive—catching operations after they've already begun rather than preventing them from starting.
The broader implication is that information warfare has entered a new phase. Foreign governments no longer need to rely solely on human operatives and traditional propaganda networks. They can now leverage AI to automate and accelerate their campaigns, making disinformation cheaper to produce and easier to distribute. As AI tools continue to improve, this asymmetry—where bad actors can move faster than platforms can respond—is likely to become more pronounced, not less.
Citações Notáveis
OpenAI's detection systems caught the operations through monitoring for suspicious activity patterns and coordinated account behavior— OpenAI's findings
Quer a matéria completa?
Leia o original em Google News ↗