OpenAI accuses China of AI influence campaign in the US

What happened?

On June 11, 2025, OpenAI published a report detailing the detection of two clusters of ChatGPT users who, allegedly from China, generated content for covert influence campaigns targeting US audiences. According to Ben Nimmo, OpenAI's principal investigator of intelligence and analysis, these users employed ChatGPT to create images and comments that were then published by batches of accounts pretending to be Americans. The promoted narratives included that building AI data centers raises electricity costs for families, and that President Donald Trump had used tariffs to keep the US ahead in the tech race. A second cluster focused on criticizing Trump's tariffs as an attempt to 'dominate tech competition.' Instructions were given in simplified Chinese and explicitly asked not to mention Chinese President Xi Jinping, which OpenAI says is an indication of the operation's Chinese origin. Both campaigns were neutralized and the accounts removed.

This incident is not an isolated case. In 2024, OpenAI had already reported removing accounts linked to influence operations from Iran and Russia that used ChatGPT to generate disinformation. However, the use of generative AI by state actors for influence campaigns has been increasing. According to a March 2025 report by cybersecurity firm Recorded Future, at least 12 countries have used AI tools in influence operations in the past year, with China being the most active. The campaign detected by OpenAI is particularly significant because it focuses on a strategic topic: AI infrastructure, an area where the US and China are fiercely competing for technological leadership.

Why is it important?

This case is relevant for several reasons. First, it shows how state actors can use generative AI tools to amplify their messages efficiently and at scale. The ability to generate personalized content in multiple formats (text, images) drastically reduces the cost and time needed to launch disinformation campaigns. Second, it highlights the vulnerability of AI systems to malicious use, even by the creators of the technology themselves. Although OpenAI has implemented safety filters, operators managed to bypass them using simplified Chinese instructions and avoiding prohibited terms. Third, the campaign sought to influence US policy on AI infrastructure, a critical issue for the country's competitiveness. According to Nimmo, the campaign amplified existing public discontent against the creation of new AI data centers, which has resulted in dozens of proposed moratoriums at local, state, and national levels. Although the actual impact was minimal — Nimmo noted that 'neither campaign appears to have gained significant authentic engagement' — it reveals the intentions and narratives that Chinese operators are testing. This occurs in a context of growing geopolitical tension over AI leadership, where the US and China are fiercely competing. In April 2025, the US government imposed new restrictions on AI chip exports to China, and Beijing responded with limits on critical mineral exports. AI-generated disinformation could become another weapon in this tech war.

Furthermore, the case underscores the difficulty of attributing influence operations. OpenAI based its conclusion on clues such as the use of simplified Chinese in prompts and the instruction not to mention Xi Jinping, but did not provide conclusive evidence of a direct link to the Chinese government. This opens the door to alternative interpretations and possible denials from Beijing. Attribution in cyberspace remains a challenge, and tech companies must be cautious when making accusations without solid evidence.

What consequences will it have?

Immediate consequences include increased scrutiny of AI use in disinformation campaigns and potential calls for stricter regulation of access to language models. US lawmakers, such as Senator Mark Warner (chairman of the Senate Intelligence Committee), have already requested hearings on the matter. In the long term, it could harden both countries' stances in the tech domain, with more export restrictions and greater investment in cybersecurity. For AI companies like OpenAI, this incident underscores the need to implement more robust safeguards and abuse detection systems. OpenAI has already announced it will improve its monitoring systems and collaborate more closely with government agencies. It could also influence public opinion and US lawmakers when approving funds for AI infrastructure, as the campaign precisely aimed to generate opposition to data centers. However, the effect could be counterproductive: by revealing the interference, it may generate greater support for AI investments as a national security measure.

On the international stage, this case could spur multilateral initiatives for AI governance. The AI Safety Summit held in Paris in February 2025 had already established principles to prevent malicious use of AI, but lacks enforcement mechanisms. Incidents like this could accelerate the creation of an international oversight body, similar to the IAEA for nuclear energy, but focused on AI. On the other hand, China could use this report to accuse the US of defamation and increase its own restrictions on foreign tech companies.

What should readers know?

Readers should be aware that AI-generated disinformation is a real and evolving threat. Although this campaign did not achieve significant traction, it sets a dangerous precedent. The ease with which fake and personalized content can be created makes it harder for the public to discern truth. It is important to verify the authenticity of viral content, especially that which stokes political or economic divisions. Tools such as source verification, metadata analysis, and consulting trusted media are essential. Tech companies must collaborate with governments and civil society organizations to develop countermeasures, such as creating automatic detection systems for AI-generated content and promoting media literacy. Finally, this case reinforces the need for international cooperation in AI governance, to prevent technology from being weaponized in geopolitical conflicts. The global community must establish clear norms and accountability mechanisms to prevent abuse, while fostering responsible innovation.