New research by CyberCX Intelligence has unearthed a network of at least 5,000 inauthentic X (formerly Twitter) accounts used to stoke political divisions in Australia, the United States, the United Kingdom and other Western democracies.
CyberCX calls the accounts the Green Cicada Network. They say they are controlled in concert by a Chinese-language artificial intelligence (AI) large language model and linked to a researcher affiliated with Tsinghua University and Zhipu AI, a prominent Chinese AI company.
Most of the accounts are currently inactive, and CyberCX says the network is likely an information operation capability in a development or experimental phase. If ramped up to full capacity, the Green Cicada Network could be used to conduct largescale disinformation operations, amplify polarising content, and undermine trust in civil and government institutions.
“This is one of the largest ever documented networks of inauthentic accounts discovered on a social media platform and could be the first significant China-related information operation run by generative AI,” said CyberCX Chief Strategy Officer Alastair MacGibbon. “While most accounts in the network are dormant, we have observed the network addressing system errors and ramping up political activity in recent months. With most of the narratives pushed by this operation relating to US politics, it’s plausible that this network is poised to interfere in the US presidential election in November. Make no mistake, this is a weapon that could be used to harm and undermine democracy in countries like Australia, the United States, and the United Kingdom.”
The network, which appears to be trained to interact with politically divisive narratives, was discovered by CyberCX Intelligence researchers in May when a cluster of X accounts began malfunctioning and producing the same AI-generated response. Accounts linked to the network have been observed writing posts or amplifying content related to:
- Criticisms of and support for US Vice President Kamala Harris;
- US electoral integrity issues, including voter ID requirements and the presidential nomination process;
- Divisive Australian domestic political issues, including nuclear energy, allegations about CFMEU corruption, and Labor’s immigration policies;
- Australia’s relationship with China; and
- The recent riots and civil unrest in the United Kingdom.
“Historically, Chinese information operations have focused on promoting pro-Chinese Communist Party ideologies,” said CyberCX Executive Director of Cyber Intelligence Katherine Mansted. “However, this approach has begun to pivot towards operations that mimic the Russian style of deepening polarisation by amplifying divisive views from multiple perspectives. The Green Cicada Network also highlights how generative AI is making scaled, malicious activity more accessible to an increasingly wide range of threat actors. It also highlights gaps in X’s willingness and ability to detect and respond to disinformation.”
CyberCX says all organisations should update their threat models to include malicious uses of generative AI and urges tech companies to take proactive steps to prevent their platforms from being exploited by threat actors using malicious AI.
