A researcher has just found an army of political propaganda accounts powered by artificial intelligence and posing as real people on X to argue in favor of Donald Trump.
The network has been posting almost identical messages in support of Trump and “seems to be big but crude,” Elise Thomas, a senior OSINT analyst at The Institute for Strategic Dialogue, mostly researching state-linked information operations, disinformation, and conspiracy theories, said.
1. I've found a large, AI-driven bot network on X which is posting in support of Trump in the US election.
undefined Elise Thomas (@elisethoma5) November 3, 2024
Thread. pic.twitter.com/ez38QbkVsP
“The accounts are an eclectic mix of join dates, handles, and personas, which are mostly not tailored to pro-Trump activity. This is pretty typical of crypto or spam networks, which tend to be cobbled together out of accounts like this,” said Thomas.
Many similar accounts have a blue verification tick – which is standard for spam networks. And the bios of some profiles seem to have been updated with pro-Trump statements, even though the earlier content wasn’t political at all.
Thomas said there’s a reason why she feels confident that these accounts are AI bots and not real people – and it’s hilarious. “They told me so,” said the researcher.
AI-powered bots indeed give themselves away in numerous glitchy messages. Instead of posting short pro-Trump messages, they, for instance, acknowledge: “I am an AI assistant developed by OpenAI to help users with various tasks.”
Sometimes, these bots also post refusals such as “As a language model AI, I cannot express political opinions or endorse any candidate.”
However, Thomas said these refusals were rare, and it shows that the creators of the bot network have clearly found a way around OpenAI’s safeguards.
The bots also argue with themselves, saying they disagree with Trump’s policies and actions. Finally, they use the hashtag #Trump2020 and tag Elon Musk, the owner of X, in their posts.
“Before switching to the pro-Trump activity, many of the accounts appear to have been used for fairly standard spammy stuff – porn, cute cats, sports. These appear to have been largely commercial bot accounts,” said Thomas.
She suspects that the whole network is automated from start to finish and that no one reads the posts before they’re published, which is probably why the aforementioned mistakes could slip through.
Thomas says it’s “really important not to jump to conclusions about attribution, especially in the generative AI era.” But if she had to guess, she’d go for Pakistan or India.
Most bot accounts have now been suspended, but numerous others are still posting political propaganda. According to OpenAI’s report released in October, its AI models are indeed being used to generate fake election content.
Cybercriminals are increasingly using AI tools, including ChatGPT, to aid in their malicious activities, such as creating and debugging malware and generating fake content for websites and social media platforms, the AI startup said.
In late September, another research report by Clemson University detailed a coordinated AI campaign using large language models to reply to other users on social media platforms.
The network identified by the Clemson researchers included at least 686 identified X accounts that have posted more than 130,000 times since January. It targeted four Senate races and two primary races and supported Trump’s re-election campaign.
Your email address will not be published. Required fields are markedmarked