OpenAI has reported the exposure of a Chinese government operation that uses artificial intelligence (AI) to monitor anti-Chinese content in real time on Western social media, The New York Times reports.
The AI-based system, which the researchers named Peer Review, was discovered after experts noticed the use of OpenAI tools to debug the code underlying this spyware.
“This is the first time we've seen such a powerful AI-based surveillance system,” said Ben Nimmo, principal researcher at OpenAI. “Threat actors sometimes give us a glimpse of what they are doing in other parts of the internet because of the way they use our A.I. models.”
According to OpenAI, the Chinese surveillance system is based on the Llama model developed by Meta. Since Llama was released as open-source, it was used to create this program.
Additionally, OpenAI uncovered another Chinese campaign called Sponsored Discontent, which used OpenAI tools to create English-language content targeting Chinese dissidents. The group also translated articles critical of the United States and distributed them in Spanish in Latin America.
The OpenAI report also highlights other threats. In particular, the company uncovered a campaign from Cambodia that used AI to participate in a fraud scheme called “pig butchering” — a scam in which men seeking online dating are targeted, then the fraudsters develop relationships with them and then lure them into investment schemes by using AI to create convincing social media comments.
OpenAI notes that, despite the risks, artificial intelligence can also be a powerful tool for detecting and preventing malicious activity.