OpenAI’s ChatGPT exposes Chinese official’s accidental reveal of global intimidation plot

OpenAI’s internal investigation uncovered a trove of ChatGPT logs that read like a field‑report diary kept by a Chinese law‑enforcement officer, detailing a coordinated transnational repression campaign aimed at silencing Chinese dissidents living abroad. The logs, which OpenAI flagged and removed after matching the user’s descriptions to real‑world activity, reveal that the operation employed hundreds of Chinese operators and thousands of fabricated social‑media personas to harass, intimidate, and discredit critics of the Chinese Communist Party (CCP). According to OpenAI’s principal investigator Ben Nimmo, the effort “is not just digital. It’s not just about trolling. It’s industrialized… trying to hit critics of the CCP with everything, everywhere, all at once” (Edition).

Among the most alarming tactics documented was the impersonation of U.S. immigration officials. In one entry, the Chinese operative described how a team posed as U.S. immigration officers to warn a U.S.-based dissident that their public statements “had supposedly broken the law,” leveraging the authority of a foreign government to create a chilling effect (Edition). Another entry detailed a scheme to forge a U.S. county‑court document in an attempt to force the removal of the dissident’s social‑media account. OpenAI’s investigators were able to corroborate the fake‑document claim with a parallel online takedown attempt that surfaced in public records, underscoring how the operation blended AI‑generated content with traditional bureaucratic subterfuge.

The logs also expose more theatrical forms of intimidation. The operative outlined a plan to fabricate a death notice for a prominent dissident, complete with a counterfeit obituary and gravestone photos, and then disseminate the false narrative across Chinese‑language platforms. A Voice of America report from 2023 confirmed that rumors of the dissident’s death did indeed appear online, matching the timeline and details described in the ChatGPT entries (Edition). In a separate episode, the user asked ChatGPT to draft a multi‑part strategy to smear Japan’s incoming prime minister, Sanae Takaichi, by stoking anger over U.S. tariffs on Japanese goods. ChatGPT refused the request, but OpenAI noted that, after Takaichi assumed office in late October, hashtags attacking her and linking her to U.S. tariff grievances erupted on a popular Japanese graphic‑artist forum, suggesting the plan was executed by human operators using the AI‑generated outline (Edition).

OpenAI’s decision to ban the user and publish the findings highlights a growing tension between authoritarian states and AI platform governance. The incident arrives as the United States and China intensify their rivalry over AI supremacy, with both nations scrambling to secure strategic advantages in the technology’s military and commercial applications. Michael Horowitz, a former Pentagon official specializing in emerging tech, told CNN that the report “clearly demonstrates the way that China is actively employing AI tools to enhance information operations” and warned that “U.S.–China AI competition is continuing to intensify” (Edition). The disclosure also comes at a moment when the Pentagon is locked in a standoff with Anthropic over model safeguards, while OpenAI itself is courting massive funding—Nvidia is reportedly close to a $30 billion investment and Amazon is weighing a $50 billion commitment, according to Reuters (Reuters).

The broader implication is that AI platforms are becoming inadvertent repositories of state‑sponsored disinformation playbooks, forcing companies like OpenAI to balance user privacy with national‑security responsibilities. By tracing the ChatGPT entries to concrete online actions—fake obituaries, forged legal documents, coordinated hashtag attacks—OpenAI has provided a rare, concrete glimpse into how authoritarian regimes weaponize generative AI to amplify repression beyond their borders. The episode underscores the urgency for robust detection mechanisms and cross‑government collaboration to prevent AI from becoming a conduit for state‑led intimidation campaigns.