A Chinese law enforcement official's use of ChatGPT accidentally exposed a sprawling influence operation targeting dissidents abroad, according to OpenAI's new report. The official used ChatGPT as a personal diary to document how Chinese operators impersonated US immigration officials and forged court documents to intimidate critics. OpenAI discovered the activity after the user detailed efforts to spread disinformation and harass Chinese dissidents living in the United States.

In one documented case, operators created fake obituaries and gravestone photos to spread rumors of a dissident's death online. The report revealed hundreds of Chinese operators managing thousands of fake social media accounts across multiple platforms. When the official asked ChatGPT to help plan attacks on Japan's incoming Prime Minister Sanae Takaichi by exploiting US tariff tensions, the AI refused the request. However, similar anti-Takaichi hashtags emerged on Japanese forums shortly after she took office.

The discovery highlights how authoritarian regimes are weaponizing AI tools for surveillance and information warfare. OpenAI banned the user after matching ChatGPT entries with real-world online activity. The incident comes amid escalating US-China competition over AI dominance, with the Pentagon pressuring companies like Anthropic over model safeguards while China expands its use of AI for transnational repression.