Key Points

According to a report in The Diplomat, the use of generative AI tools to tailor content to local languages and cultural contexts alongside a focus on youth could leverage social media's popularity to deceptively build trust in pro-Beijing sources and influence future leaders in developing regions.

The report also highlights that in early August, two professors from Vanderbilt University in the US published an essay outlining a trove of Chinese documents linked to the private firm GoLaxy. The sources revealed that artificial intelligence (AI) was being used to generate misleading content for target audiences, such as in Hong Kong and Taiwan, and also to extract information about US lawmakers, creating profiles for possible use in future espionage operations or influence campaigns.

The report states that there have also been several incidents involving OpenAI, Meta, and Graphika on the latest uses of AI by China-linked actors focused on foreign propaganda and disinformation. This calls for urgent attention from social media platforms, software developers, and democratic governments, it added.

According to the report, while prior China-linked disinformation campaigns had deployed AI tools to generate false personas or deepfakes, these latest disclosures point to a more concerted effort to leverage these tools for creating entire fake news websites that distribute Beijing-aligned narratives simultaneously in multiple languages. Graphika's "Falsos Amigos" report, published last month, identified a network of 11 fake websites, established between late December 2024 and March 2025, using AI-generated pictures as logos or cover images to enhance credibility.

OpenAI's threat report published in June cited the use of similar tactics, noting that now-banned ChatGPT accounts had used prompts (often in Chinese) to generate names and profile pictures for two pages posing as news outlets, as well as for individual persona accounts of US veterans critical of the Trump administration in a campaign the firm dubbed "Uncle Spam". These efforts aimed to fuel political polarisation in the United States, with AI-crafted logos and profiles amplifying the illusion of authenticity.

Another key strategy involved simulating organic engagement. OpenAI detected China-linked accounts bulk-generating social media posts, with a "main" account posting a comment followed by replies from others to mimic a discussion. The "Uncle Spam" operation generated comments from supposed American users both supporting and criticising US tariffs.

It highlighted the case of Pakistani activist Mahrang Baloch, who has criticised China's investments in the disputed territory of Balochistan. Meta documented a TikTok account and Facebook page posting a false video accusing her of appearing in pornography, followed by hundreds of apparently AI-generated comments in English and Urdu to simulate responses.

- IANS