Using American companies' own tools to manufacture dissent
In the ongoing contest between nations over technological and economic dominance, China's deployment of ChatGPT to manufacture American dissent marks a quiet but consequential threshold: the tools of democratic discourse have become instruments of its subversion. OpenAI disclosed that Chinese operatives used the platform to craft sophisticated propaganda targeting wealthy stakeholders and policymakers on questions of AI infrastructure and trade tariffs. The campaign did not announce itself as foreign interference — it wore the grammar and cadence of domestic debate. What emerges is a sobering reminder that the accessibility which makes AI powerful for millions also makes it available to those who would hollow out the conversations it enters.
- Chinese operatives ran a sustained, coordinated campaign using ChatGPT to generate polished, contextually convincing arguments against US data center expansion and tariff policy — content designed to pass as organic American dissent.
- The targeting was deliberate and precise: wealthy individuals and Washington-adjacent decision-makers, the constituencies whose opposition carries disproportionate weight in shaping policy outcomes.
- OpenAI's own monitoring systems eventually flagged and shut down the accounts, but not before the campaign had operated long enough to raise hard questions about how much damage had already been done.
- The incident tears open a structural vulnerability — AI platforms built for openness and helpfulness have no natural immune response to state-level actors who exploit those very qualities.
- The disclosure lands at a moment of acute sensitivity, with the US actively navigating both data center policy and tariff strategy as pillars of its broader technological rivalry with China.
- The barrier to sophisticated disinformation has effectively collapsed: a small, resourced team can now produce and distribute tailored influence content at scale, using tools that require no special access to obtain.
OpenAI revealed that operatives based in China had used ChatGPT in a coordinated campaign to generate propaganda targeting American policy debates around AI infrastructure and trade tariffs. The effort was not crude or easily identifiable as foreign in origin — the actors used the AI system to produce sophisticated, contextually fluent arguments designed to resonate with American audiences and policy circles, then distributed them through networks of accounts to simulate grassroots opposition.
What distinguished the campaign was its strategic focus. Rather than broadcasting broadly, the operatives directed their messaging at wealthy individuals and influential decision-makers — people positioned to exert real pressure on Washington. By manufacturing doubt about data center expansion and tariff policy among this audience, the actors sought to move levers that typically respond to elite opinion. It was information warfare recalibrated for the AI era: using an American company's own product to serve Chinese strategic interests.
OpenAI's security systems eventually identified and shut down the accounts, but the timeline of the campaign exposed a troubling gap between the sophistication of state-sponsored actors and the detection mechanisms meant to catch them. The company's usage policies had not prevented a sustained operation from taking root.
The broader implication cuts deeper than any single campaign. State actors no longer need to build propaganda infrastructure from the ground up — they can access the most capable language tools in the world, often openly, and produce tailored influence content at scale. The same qualities that make AI systems useful to millions make them exploitable by a determined few. For OpenAI and the industry it leads, the incident posed a question that existing safeguards had not yet answered: how do you build a tool for everyone without building one for anyone?
OpenAI disclosed that operatives based in China had systematically used ChatGPT to generate and distribute propaganda aimed at shaping American policy debates around artificial intelligence infrastructure and trade tariffs. The campaign, discovered through the company's security monitoring, represented a coordinated effort to manufacture opposition to US data center expansion among wealthy stakeholders and policymakers who might influence regulatory decisions.
The operatives deployed ChatGPT—the same tool millions of Americans use daily for writing, coding, and research—to produce persuasive content that opposed both the construction of new AI data centers on American soil and the tariff policies the US government had begun implementing. Rather than creating crude or obviously foreign messaging, the actors used the AI system to generate sophisticated, contextually appropriate arguments designed to resonate with American audiences and policy circles. The content was then distributed through coordinated networks of accounts, amplifying the reach and apparent grassroots nature of the opposition.
What made this campaign notable was its targeting strategy. The operatives focused their messaging toward wealthy individuals and decision-makers—people with the resources and influence to shape policy conversations. By seeding doubt about data center expansion and tariff policies among this audience, the actors aimed to create pressure on policymakers from constituencies that typically have outsized influence in Washington. The campaign represented a form of information warfare adapted for the AI age: using American companies' own tools to manufacture dissent on issues that benefited Chinese interests.
The discovery raised immediate questions about the safeguards built into large language models and the platforms that deploy them. OpenAI's systems had allowed the accounts to operate long enough to conduct a sustained campaign, suggesting that detection mechanisms for state-sponsored manipulation may lag behind the sophistication of the actors deploying them. The company moved to shut down the accounts once the activity was identified, but the incident exposed a vulnerability: AI systems designed to be helpful and accessible could be weaponized by foreign governments to influence American political and economic decisions.
The timing of the disclosure was significant. The US had been actively debating both data center policy and tariff strategy as part of its broader competition with China over technological dominance. Data centers consume enormous amounts of electricity and require substantial infrastructure investment, making them politically contentious in communities where they might be built. Tariffs, meanwhile, directly affect American businesses and consumers, making them perpetually contested terrain in policy debates. By targeting both issues simultaneously, the Chinese operatives were attempting to influence multiple pressure points in American decision-making.
The incident illustrated a broader pattern emerging in the information warfare landscape: state actors no longer needed to create their own propaganda from scratch. They could simply use the most advanced AI tools available, often legally and openly, to generate the content they needed. The barrier to entry for sophisticated disinformation campaigns had collapsed. A small team with access to ChatGPT and basic social media infrastructure could potentially reach millions of Americans with tailored messaging designed to serve foreign policy objectives.
OpenAI's disclosure also highlighted the company's own position in this ecosystem. As the creator of one of the world's most powerful language models, OpenAI bore some responsibility for how the tool was being used. The company had implemented usage policies and monitoring systems, but the Chinese campaign demonstrated that determined state actors could still find ways to exploit the platform. The question facing OpenAI and other AI companies was whether existing safeguards were sufficient, or whether new approaches to detecting and preventing state-sponsored manipulation would be necessary.
Notable Quotes
OpenAI disclosed that operatives based in China had systematically used ChatGPT to generate and distribute propaganda aimed at shaping American policy debates— OpenAI security findings
The Hearth Conversation Another angle on the story
Why would China care enough about US data centers to run a propaganda campaign?
Data centers are where AI models live and train. If the US builds more of them domestically, it accelerates American AI development and reduces dependence on Chinese chips and infrastructure. That's a direct threat to China's technological position.
But couldn't they just say that directly? Why use ChatGPT to manufacture opposition?
Because a Chinese government statement saying "don't build data centers" gets dismissed immediately. But if wealthy Americans start arguing against data centers for reasons that sound homegrown—environmental concerns, local disruption, economic logic—that carries weight with policymakers.
So they were essentially ghostwriting arguments that Americans would find convincing?
Exactly. ChatGPT can generate contextually appropriate, persuasive content in American English that sounds like it came from someone who lives here and understands the issues. That's far more effective than crude propaganda.
How did OpenAI even catch this?
They monitor for coordinated inauthentic behavior—patterns of accounts that don't look like real people, that operate in suspicious ways. But the sophistication of the content itself probably made detection harder. The arguments weren't obviously false; they were just strategically chosen.
What happens now? Can they just do this again with a different tool?
Probably. The vulnerability isn't ChatGPT specifically—it's that AI systems are now powerful enough to generate convincing propaganda at scale, and detection lags behind deployment. Until companies build better safeguards, this will keep happening.