OpenAI Uncovers Industrial-Scale Chinese Influence Campaign Using ChatGPT

OpenAI has revealed a sophisticated, state-aligned influence campaign that leveraged its ChatGPT platform to produce disinformation, manipulate public discourse, and intimidate critics on a global scale. The operation, uncovered through internal monitoring systems, was initially flagged by anomalous usage patterns tied to a single Chinese government official who repeatedly requested AI-generated content designed to harass overseas dissidents and smear political opponents. What began as an isolated anomaly soon unraveled into a vast network of coordinated AI-driven propaganda, spanning over 20 countries and involving hundreds of fabricated personas and fake news outlets.

According to OpenAI’s Threat Intelligence Team, the campaign utilized ChatGPT to generate thousands of posts, comments, and articles in multiple languages—primarily English, Spanish, and Southeast Asian dialects—targeting human rights activists, journalists, and expatriate communities critical of Beijing’s policies. The AI-generated content was designed to mimic organic social media discourse, often referencing real events while inserting misleading narratives, such as falsely portraying protesters as foreign-funded agitators or fabricating scandals involving dissident figures.

The investigation traced the operation’s origins to a Chinese government-affiliated research institute that had gained access to ChatGPT through a legitimate enterprise subscription. Rather than using the tool for legitimate research, the entity systematically trained its staff to craft highly specific prompts optimized to produce politically charged, emotionally manipulative outputs. These outputs were then distributed across forums, Reddit threads, Facebook groups, and even local news comment sections in target countries, creating the illusion of grassroots opposition to dissident voices.

Notably, the operation did not rely solely on automated bots. Human operators reviewed and curated each AI-generated output, adding local slang, cultural references, and regional political context to enhance credibility. In one case, an AI-generated article falsely accused a Uyghur activist in Germany of terrorism, which was then republished by a small, obscure blog in Malaysia and later cited by a state-aligned news outlet in Pakistan. OpenAI confirmed that over 12,000 unique pieces of content were generated during a six-month period, with a significant portion appearing on platforms with minimal moderation.

OpenAI has since suspended the offending enterprise account and notified relevant governments and international organizations, including the U.S. State Department and the European Union’s Disinformation Task Force. The company has also updated its content moderation protocols to detect prompt patterns associated with state-sponsored influence operations, including repeated requests for defamatory content, geopolitical manipulation, and identity fabrication.

While OpenAI emphasized that ChatGPT itself was not compromised—its systems were used as a tool, not a vector—the incident underscores the growing threat of AI-enabled hybrid warfare. Experts warn that such campaigns, if left unchecked, could erode trust in digital discourse and destabilize democratic institutions by amplifying division under the guise of public opinion.

China’s Ministry of Foreign Affairs has not formally responded to the allegations. However, anonymous sources within Beijing’s cyber diplomacy apparatus told Reuters that China opposes any "unfounded accusations" against its institutions and emphasized its commitment to "responsible AI development." Meanwhile, OpenAI has pledged to increase transparency around enterprise account usage and is collaborating with academic researchers to develop AI watermarking techniques that could help identify state-sponsored synthetic content in real time.

This case marks one of the first documented instances of a nation-state deploying generative AI at industrial scale for psychological operations. As AI tools become more accessible, the line between private innovation and state espionage grows increasingly blurred—and the world’s tech platforms are now on the front lines of a new kind of information war.