- 14 Mar 2002
- 21,056
- 18,911
The AI war is on:
OpenAI said it banned a ChatGPT account linked to an individual affiliated with Chinese law enforcement after the user tried to use the system to plan a covert influence campaign targeting Japanese Prime Minister Sanae Takaichi. According to a report released Monday, titled "Disrupting Malicious Uses of AI," ChatGPT refused the original request. The account then turned to AI models developed by DeepSeek and Alibaba, including Qwen, to draft a plan. The material was later brought back to ChatGPT for refinement.
The draft outlined steps to amplify online negative commentary about Takaichi. It also proposed creating fake email accounts posing as foreign residents to send complaints about her to Japanese politicians. The plan included efforts to stir online anger over U.S. tariffs and to use Japan's ties with Washington to divert attention from relations with China.
OpenAI said some China-linked users also asked ChatGPT to help draft posts aimed at Chinese dissenters, both inside and outside the country, targeting them online and offline. The company said available evidence points to what it described as "cyber special operations" carried out by Chinese law enforcement. The activity appeared large in scale and sustained, involving hundreds of personnel, thousands of fake accounts across numerous platforms, locally deployed AI models, and a range of operational tactics. OpenAI's services are not available in mainland China or Hong Kong.
Paywall alert:
asia.nikkei.com
OpenAI said it banned a ChatGPT account linked to an individual affiliated with Chinese law enforcement after the user tried to use the system to plan a covert influence campaign targeting Japanese Prime Minister Sanae Takaichi. According to a report released Monday, titled "Disrupting Malicious Uses of AI," ChatGPT refused the original request. The account then turned to AI models developed by DeepSeek and Alibaba, including Qwen, to draft a plan. The material was later brought back to ChatGPT for refinement.
The draft outlined steps to amplify online negative commentary about Takaichi. It also proposed creating fake email accounts posing as foreign residents to send complaints about her to Japanese politicians. The plan included efforts to stir online anger over U.S. tariffs and to use Japan's ties with Washington to divert attention from relations with China.
OpenAI said some China-linked users also asked ChatGPT to help draft posts aimed at Chinese dissenters, both inside and outside the country, targeting them online and offline. The company said available evidence points to what it described as "cyber special operations" carried out by Chinese law enforcement. The activity appeared large in scale and sustained, involving hundreds of personnel, thousands of fake accounts across numerous platforms, locally deployed AI models, and a range of operational tactics. OpenAI's services are not available in mainland China or Hong Kong.
Paywall alert:
OpenAI flags China-linked influence ops targeting Japan's Takaichi
Attempt went ahead with DeepSeek and Alibaba's Qwen after ChatGPT denied request
asia.nikkei.com
