OpenAI Bans China-Linked Accounts Using ChatGPT for U.S. Influence Campaigns

OpenAI has banned accounts linked to Chinese operatives that used ChatGPT to generate social media content targeting U.S. debates over AI data centers and tariffs. The operation marks the first time OpenAI has identified a China-linked campaign explicitly meddling in the AI data center debate, though the company says it failed to gain significant traction.

• What Happened
• The Two Campaigns
• Why It Matters
• Market and Policy Implications
• What's Next
• Frequently Asked Questions
• Conclusion

What Happened

OpenAI disclosed Wednesday that it had uncovered and disrupted two influence operations originating from China that relied on ChatGPT to draft posts, comments, and political cartoons. The company attributed the activity to accounts it believes were linked to a Chinese government contractor and a separate, unattributed group. Both operations targeted heated U.S. political debates around technology infrastructure and trade policy.

“This was not a case of an influence operation creating a debate,” Ben Nimmo, principal investigator on OpenAI’s intelligence and investigations team, told reporters. “The debate existed already. This was an influence operation from China trying to interfere in it.”

According to Axios, OpenAI said the campaigns did not appear to have achieved meaningful reach or engagement online.

The Two Campaigns

OpenAI detailed two distinct operations, each targeting a different U.S. policy flashpoint.

Data Center Bandwagon: This campaign generated comments and comic strips portraying AI data centers as drivers of rising electricity prices for American households. Users asked ChatGPT to create cartoons about power grid capacity, then posted them to X through likely inauthentic accounts alongside links to legitimate news coverage of data center energy demand. The operation attempted to amplify existing public concerns: a recent Harvard/MIT poll found 32% of Americans oppose data centers in their area, while 40% support them.

Tech and Tariffs: A second operation used ChatGPT to produce content and political cartoons criticizing Trump-era tariffs and the U.S. push for global technology dominance. In one example, OpenAI described a cartoon depicting President Trump wearing American flag pants labeled “America First,” holding a mallet inscribed with “Tech Dominance” and swinging it into a wall reading “Global Future.” The campaign aligned with existing public sentiment: a Harris poll released in March found seven in 10 Americans said tariffs have caused them to pay higher prices.

Why It Matters

The incident highlights a new frontier in foreign influence operations: the use of generative AI to cheaply scale content production around existing political divisions. Unlike traditional disinformation campaigns that rely on human content creation or simple bot networks, these operations leverage large language models to produce varied, context-aware material — including images — that can be rapidly deployed across social platforms.

OpenAI said this appears to be the first time the company has seen a China-linked operation using its models to meddle in the AI data center debate specifically. The timing is notable: data center construction is booming, driven by cloud providers and AI startups, while local opposition over energy use and environmental impact is growing. Foreign actors seeking to amplify that opposition now have a powerful new tool.

The campaigns also illustrate a broader shift. Last year, researchers documented how Chinese state-linked actors used AI tools to create realistic-looking news sites and social media personas, as covered in Axios’s earlier report on AI-powered disinformation. The latest incident suggests influence operators are increasingly layering generative AI onto existing playbooks.

Market and Policy Implications

For technology companies, the incident underscores the growing burden of content moderation and adversarial threat detection. OpenAI’s ability to detect and ban these accounts — and to publicly attribute the activity — is itself a signal of escalating counter-measures. But the low cost of generating AI content means that even failed campaigns can serve as stress tests, forcing platforms and AI providers to constantly update their defenses.

For investors and policymakers, the news adds urgency to debates over AI governance and export controls. The U.S. government has already tightened restrictions on advanced chip sales to China, partly to limit the development of AI capabilities that could be used for disinformation. This incident may reinforce calls for more aggressive enforcement and for closer collaboration between AI companies and intelligence agencies.

Meanwhile, public opinion on data centers remains a live issue. With billions of dollars in planned infrastructure at stake — from hyperscale data centers to edge facilities — even a small shift in local sentiment could affect permitting, tax incentives, and site selection. Foreign influence campaigns that successfully inflame opposition could disrupt supply chains and raise costs for builders and operators.

What's Next

OpenAI said it will continue to monitor for similar operations and share threat intelligence with industry partners and governments. The company has also updated its usage policies and enforcement systems to better detect coordinated influence campaigns that use its models.

The broader industry may face new regulatory scrutiny. Congress has shown interest in requiring AI companies to report foreign misuse, and the Biden administration previously issued an executive order on AI safety that included provisions on disinformation. Whether the next administration continues those efforts remains uncertain.

For now, the episode serves as a warning: generative AI is making it easier and cheaper for adversarial actors to enter U.S. political debates — and the debates themselves are growing more heated.

Frequently Asked Questions

How did OpenAI detect these campaigns? OpenAI’s intelligence and investigations team analyzed account behavior, content patterns, and cross-referenced known indicators of state-linked activity. The company said it identified the campaigns through a combination of automated monitoring and manual investigation.

Were any U.S. politicians or media targeted? The campaigns targeted general public debate rather than specific individuals or outlets. They posted comments and cartoons on social media platforms, often linking to legitimate news stories to create a veneer of credibility.

Could these campaigns have been more effective with different tactics? OpenAI noted the operations failed to gain significant traction, likely because the content was generic and the accounts lacked organic reach. A more sophisticated operation — using tailored language, multiple languages, or coordinated with real human influencers — could potentially achieve greater impact.

What can other AI companies learn from this incident? The case highlights the importance of investing in threat detection, maintaining clear usage policies, and sharing intelligence across the industry. It also shows that even ineffective campaigns provide valuable data for refining models and enforcement.

Does this affect OpenAI’s business relationships with China? ChatGPT is not officially available in China, but some users access it via VPNs. OpenAI has previously restricted API access in certain regions. The company said it will continue to enforce its policies against misuse regardless of geography.

How does this compare to previous Chinese disinformation campaigns? Traditional campaigns relied on human content creators or bot networks with repetitive messaging. AI-generated content allows for more variety — different arguments, images, and formats — at a fraction of the cost, making it harder to detect and counter.

Conclusion

OpenAI’s disclosure of two China-linked influence campaigns represents a notable escalation in the use of generative AI for political interference. While neither operation gained real-world traction, the fact that state-backed actors are now testing ChatGPT to intervene in U.S. data center and tariff debates signals a new phase in the disinformation arms race. As AI tools become cheaper and more capable, both companies and governments will need to adapt rapidly to protect public discourse from manipulation.