Thousands of AI-controlled social media accounts designed to disrupt the U.S. election have been linked to China, according to a report by cyber investigators.

The network, known as “Green Cicada”, has been uncovered by cyber security firm CyberCX, which found over 5,000 accounts on social media platform X.

The Green Cicada network operates around the clock, generating and amplifying politically charged content across multiple time zones.

The accounts use AI technology to reword existing tweets and news articles, with the aim of creating confusion and division. The accounts are not only focusing on the U.S., as the operation is also targeting the UK, Western Europe, and Australia.

CyberCX’s investigation traced the network back to a Chinese university and an AI company, raising concerns about state-sponsored disinformation efforts.

The firm’s experts believe that this may be one of the largest publicly identified networks of inauthentic X profiles, warning that it may feature more that 8,000 accounts.

CyberCX also suggest that it could represent the first major China-related information operation that heavily relies on generative AI technology.

A senior intelligence source speaking to National Security News on the grounds of anonymity said: “This report clearly demonstrates that the Chinese government wants to use covert social media accounts to interfere with the US elections. One can only assume that the reasons behind this are that China wants the next administration to be damaging to the US but beneficial to its own interests.”

CyberCX noted that many of the accounts in the Green Cicada network appear to be dormant, so they are not activity posting content. The firm said that this suggests that they may be part of a “sleeping giant” ready to become more active as the U.S. election in November approaches.

The accounts in the network have been increasing their activity since around April 2024.

In August, the firm found a network campaign that focused on posting comments relating to the integrity of U.S. elections and the presidential candidate nomination process. Other campaigns focused on retweets criticising US Vice President Kamala Harris, while another focused on comments in support of her.

Some posts begin with phrases like “As a Spanish citizen…” or “As a fan of the U.S. political system from Germany…”, giving the appearance of diverse, global perspectives.

“Amplification of divisive content in democracies is consistent with China’s information operations playbook,” CyberCX stated, highlighting the broader strategy behind Green Cicada’s activities.

While primarily targeting the United States, the network has also been active in other countries. In the UK, for example, the Green Cicada accounts have amplified narratives around immigration, support for Israel, and even linked local incidents, such as the Southport stabbings on 29 July, to illegal migration.

The tell-tale signs of the accounts being AI-powered include the use of AI-generated profile pictures, where pupil alignment is identical across images.

Additionally, some accounts were identified when they posted AI refusal text, stating, “As an AI language model, I must follow the guidelines and principles set by the Chinese government. The content you provided contains sensitive topics that may lead to discrimination and hatred.”

CyberCX state that these comments suggest that the AI was prompted to harmfully reword content across a range of topic. The firm observed topics which included U.S. First Lady Dr Jill Biden, an Iranian religious leader, Indian domestic politics, and high-profile U.S. lobbyists.

The discovery of the Green Cicada network comes amidst broader concerns over the use of AI in election interference. Microsoft recently reported a surge in Iranian influence campaigns and noted that nearly all actors involved in such activities are now incorporating AI-generated content into their operations.

Despite these findings, the U.S. Office of the Director of National Intelligence (ODNI) recently stated that China “probably does not plan to influence the outcome of the U.S. presidential election.” However, the ODNI did acknowledge that Chinese influence actors “are using social media to sow divisions in the United States and portray democracies as chaotic.”