OpenAI Exposes Chinese Law Enforcement Use of ChatGPT in Covert Smear Campaigns

OpenAI Exposes Chinese Law Enforcement Use of ChatGPT in Covert Smear Campaigns

OpenAI has revealed a disturbing new pattern of AI misuse: Chinese law enforcement personnel reportedly leveraged ChatGPT to plan and monitor digital smear campaigns against international political figures, including Japan’s Prime Minister and other critics of the Chinese Communist Party (CCP). According to OpenAI’s latest report on adversarial AI use, the malicious activity involved generating fabricated narratives, coordinating social media disinformation, and analyzing public sentiment to refine targeting strategies—all through prompts issued to its language model.

The findings, first reported by MSN and corroborated by The Register, detail how a network of accounts with apparent ties to Chinese state security agencies utilized ChatGPT to produce convincing, culturally tailored content designed to discredit opponents. These included fabricated quotes, doctored images, and misleading articles posted across forums, Twitter/X, and Chinese social media platforms under false identities.

OpenAI’s security team flagged the activity in late 2025 after detecting anomalous usage patterns: unusually high volumes of politically charged prompts, repeated requests for tone modulation to mimic Western media styles, and attempts to bypass content moderation filters by using indirect phrasing. One particularly concerning example involved an operator asking ChatGPT to generate a series of tweets portraying the Japanese Prime Minister as “corrupt and pro-Western,” then using the output to seed trending hashtags in Japanese and English-speaking online communities.

Unlike typical cybercriminals who use AI for phishing or spam, this case represents a sophisticated, state-aligned operation aimed at influencing foreign public opinion. The actors reportedly used ChatGPT not just as a content generator but as an intelligence tool—querying it to assess the likely public reaction to various narratives, simulate media coverage, and even draft counter-narratives to neutralize dissenting voices abroad.

OpenAI has since suspended the identified accounts and updated its content moderation systems to better detect state-sponsored manipulation attempts. However, the incident underscores a growing global challenge: as AI becomes more accessible, authoritarian regimes are rapidly integrating it into their information warfare arsenals. “This isn’t just misuse—it’s weaponization,” said an OpenAI spokesperson, speaking on condition of anonymity. “We’re seeing state actors treat our models like digital field manuals for psychological operations.”

Experts warn that such operations could set a dangerous precedent. “If one nation openly uses AI to manipulate foreign discourse without accountability, others will follow,” said Dr. Elena Ruiz, a disinformation researcher at Stanford’s Cyber Policy Center. “The line between AI-assisted propaganda and cyber espionage is blurring—and international norms haven’t caught up.”

China’s Ministry of Foreign Affairs has not publicly commented on the allegations. However, Chinese state media have previously dismissed similar accusations as “Western misinformation,” framing AI regulation as a matter of national sovereignty. Meanwhile, OpenAI has called on global tech platforms and governments to establish multilateral frameworks for monitoring AI-driven influence operations, particularly those involving state actors.

The revelation comes amid heightened scrutiny of AI’s role in global politics, following similar reports of AI-generated deepfakes in Ukraine and election interference attempts in Latin America. As AI models grow more powerful and accessible, the need for transparent oversight, ethical guardrails, and cross-border cooperation has never been more urgent.

For now, OpenAI continues to monitor for signs of similar activity and has pledged to share threat intelligence with international cybersecurity agencies. But as this case demonstrates, the most dangerous use of AI may not be in hacking systems—it’s in hacking truth itself.