A remarkable intelligence failure has laid bare one of Beijing’s most sensitive covert operations: Chinese government officials apparently used OpenAI’s ChatGPT to draft threatening messages, craft surveillance reports, and generate propaganda targeting Chinese dissidents living abroad. The discovery, made possible by the AI company’s own threat intelligence team, has provided an unprecedented window into how authoritarian states are adapting commercial artificial intelligence tools for repression that extends far beyond their own borders.
The revelations, first reported by Slashdot, stem from OpenAI’s ongoing efforts to identify and shut down state-linked accounts that misuse its platform. According to the findings, Chinese officials used the AI chatbot to compose messages intended to intimidate overseas critics of the Chinese Communist Party, translate surveillance documents, and generate social media content designed to discredit prominent dissidents. The operation, which researchers have linked to China’s broader transnational repression apparatus, demonstrates how widely available AI tools are being co-opted for authoritarian purposes with startling efficiency.
AI as a Force Multiplier for State Intimidation
OpenAI’s threat intelligence report detailed how the accounts in question used ChatGPT to polish threatening letters written in English and other languages, suggesting the operators were not native speakers of the languages in which they needed to communicate. In several cases, the prompts fed into the system explicitly referenced known Chinese dissident figures and requested language designed to discourage their activism. The AI was also used to summarize social media posts by these individuals, effectively turning the chatbot into a monitoring tool that could rapidly process large volumes of open-source information about targets.
The operation aligns with what human rights organizations have long documented as China’s sprawling transnational repression network. Freedom House, which tracks such activities globally, has identified China as the world’s most prolific practitioner of transnational repression, targeting exiled Uyghurs, Tibetans, Hong Kong democracy activists, and Falun Gong practitioners across dozens of countries. What makes this latest discovery different is the technological dimension: the use of a commercially available, American-made AI system to facilitate the work.
The Mechanics of Digital Repression
According to reporting by The New York Times and other outlets that have covered China’s overseas influence operations extensively, Beijing’s security services have traditionally relied on human operatives — sometimes deployed through Chinese consulates, sometimes recruited from diaspora communities — to monitor and pressure dissidents abroad. These operations have included physical surveillance, harassment of family members still in China, and cyberattacks against activists’ devices. The integration of AI chatbots into this toolkit represents a significant operational evolution, one that allows a smaller number of operatives to produce higher volumes of targeted content across multiple languages.
OpenAI’s report indicated that the accounts were shut down after the company’s systems flagged patterns of misuse. The company has previously disclosed similar state-linked operations originating from Russia, Iran, and North Korea, but the Chinese campaign stood out for its focus on individual dissidents rather than broader disinformation themes. The prompts recovered from the accounts painted a picture of a systematic effort: operators would input a dissident’s name, request background information, then ask the AI to draft messages calibrated to instill fear or undermine the target’s credibility within their community.
A Growing Pattern of Authoritarian AI Adoption
This is not the first time OpenAI has uncovered foreign government misuse of its platform. In its February 2024 threat report, the company disclosed that it had disrupted five state-affiliated influence operations, including campaigns linked to China, Russia, and Iran. Those earlier operations focused primarily on generating social media propaganda and fake news articles. The latest Chinese operation, however, marks an escalation — moving from broad influence campaigns to targeted harassment of specific individuals, a form of AI-enabled persecution that raises urgent questions about the responsibilities of AI companies and the governments that regulate them.
Ben Wizner, director of the ACLU’s Speech, Privacy, and Technology Project, has noted in previous commentary that the use of AI for targeted repression creates asymmetries that are difficult for individual victims to counter. A single government operative armed with a chatbot can now produce personalized threats in dozens of languages, monitor targets’ social media output in real time, and generate convincing disinformation about specific people — capabilities that previously would have required teams of translators, analysts, and propagandists.
The Diplomatic and Legal Fallout
The exposure of this operation arrives at a particularly fraught moment in U.S.-China relations. Washington has already imposed restrictions on the export of advanced AI chips to China and has pressured allies to follow suit. Beijing, for its part, has accelerated development of domestic AI alternatives such as DeepSeek, which burst onto the global stage in early 2025 with capabilities that surprised Western researchers. The revelation that Chinese officials were using an American AI product for repressive purposes could further harden the stance of lawmakers who argue for tighter controls on AI technology transfers.
On Capitol Hill, the discovery has drawn attention from members of both parties. The House Select Committee on the Chinese Communist Party, which has held multiple hearings on transnational repression, has previously called for stronger enforcement actions against Chinese operatives working on American soil. The use of ChatGPT in these operations adds a new dimension to those concerns, as it demonstrates that even without physical presence in the United States, foreign agents can exploit American technology infrastructure to carry out repression remotely.
OpenAI’s Expanding Role as a Counterintelligence Actor
OpenAI’s willingness to publicly disclose state-linked threats has positioned the company in an unusual role — part technology provider, part counterintelligence outfit. The company’s threat intelligence team, which includes former national security professionals, has built detection systems designed to identify coordinated inauthentic behavior on its platform. These systems look for patterns such as accounts that consistently generate content about politically sensitive topics, use prompts that reference specific intelligence targets, or exhibit usage patterns consistent with shift work at government agencies.
Yet the company’s ability to detect and disrupt these operations is inherently limited. OpenAI can shut down accounts, but it cannot prevent operatives from creating new ones using different credentials. The company also faces the challenge of balancing transparency with operational security — disclosing too much about its detection methods could help adversaries evade them. In its public reports, OpenAI has been careful to describe the nature of the threats it has identified without revealing the specific technical indicators that led to detection.
Implications for Dissidents and Democratic Governments
For the dissidents targeted by these operations, the revelations are both vindicating and alarming. Many Chinese activists abroad have long reported receiving threatening messages that seemed unusually polished and well-informed, suspecting government involvement but lacking proof. The OpenAI disclosures provide a degree of confirmation, but they also underscore the scale of the threat. If Chinese security services are using ChatGPT — a tool whose usage can be tracked and disrupted — they are almost certainly also using domestic Chinese AI systems that operate beyond the reach of Western companies and regulators.
Safeguard Defenders, a Madrid-based human rights organization that has extensively documented China’s overseas police stations and intimidation campaigns, has warned that AI-powered repression could make it significantly harder for dissidents to distinguish genuine threats from automated harassment, and could overwhelm the limited resources of organizations that support targeted individuals. The group’s research has shown that Chinese transnational repression operations have been documented in at least 53 countries, affecting hundreds of thousands of people.
What Comes Next for AI Governance and Human Rights
The incident raises fundamental questions about the governance of powerful AI systems in an era of intensifying geopolitical competition. Current U.S. regulations do not specifically prohibit foreign governments from using American AI products, though such use may violate OpenAI’s terms of service and potentially run afoul of sanctions and export control laws depending on the specific entities involved. The European Union’s AI Act, which began taking effect in stages in 2025, includes provisions addressing AI systems used for social scoring and mass surveillance, but its applicability to transnational repression conducted via commercial chatbots remains untested.
For now, the burden of detection and enforcement falls primarily on AI companies themselves — an arrangement that critics argue is insufficient given the stakes involved. As AI systems grow more capable and more widely accessible, the potential for their misuse by authoritarian governments will only increase. The Chinese ChatGPT operation may be among the first documented cases of AI-enabled transnational repression, but it is unlikely to be the last. The question facing policymakers, technologists, and human rights advocates alike is whether governance frameworks can evolve quickly enough to address a threat that is already here.
WebProNews is an iEntry Publication