OpenAI has reportedly banned several accounts that were associated with a Chinese surveillance tool. These accounts were using ChatGPT to create sales pitches and debug code for an AI assistant to gather real-time data on anti-China protests occurring in Western countries, like the U.S. and the U.K.
According to a report from Bloomberg, OpenAI's principal investigator, Ben Nuimmo, said that the company is trying to shed light on how authoritarian regimes like China can try to leverage U.S.-built tech against the U.S. and its allies.
“This is a pretty troubling glimpse into the way one non-democratic actor tried to use democratic or US-based AI for non-democratic purposes, according to the materials they were generating themselves.”
OpenAI further said that these accounts had also referenced other AI tools, including a version of Llama, which is the open-source AI model developed by Meta. From what OpenAI could investigate, it seems like the surveillance software was called “Qianyue Overseas Public Opinion AI Assistant", although it wasn't verified whether it was deployed or not.
In its report, OpenAI said that its policies include preventing the use of AI tools by authoritarian regimes to amass power and control their citizens.
"We believe that making sure AI benefits the most people possible means enabling AI through common-sense rules aimed at protecting people from actual harms, and building democratic AI.
This includes preventing use of AI tools by authoritarian regimes to amass power and control their citizens, or to threaten or coerce other states; as well as activities such as covert influence operations (IOs), child exploitation, scams, spam, and malicious cyber activity."
The software was created to identify online threads and conversations in Western countries about human rights in China and other anti-China topics, from social media websites such as X, Facebook, and Instagram, and to send these surveillance reports to Chinese authorities, intelligence agents, and staff at Chinese embassies.
OpenAI has been warning the U.S. government for a long time, especially after DeepSeek stole the show with its R1 reasoning model which surpassed many other U.S.-made models in benchmarks and caused havoc in the U.S. markets. OpenAI accused DeepSeek of distilling the output of its models, and that Chinese companies could be stealing proprietary data from the U.S. companies.
Source: Bloomberg
0 Comments - Add comment