'For good and for evil': Study reveals AI army flooding X with GOP propaganda
An artificial intelligence army has quietly flooded social media with rightwing propaganda in tens of thousands of posts revealed in a new report.
A coordinated AI campaign with murky origins has used ChatGPT-like technology to engage with X users since March, according to a new Clemson University analysis.
"The profile images include a range of themes," the report states. "They often employ simple cartoon characters but also frequently include imagery suggestive of conservative values; many include images of former President [Donald] Trump."
Clemson University researchers found 686 X accounts had published 130,000 posts about the 2024 presidential campaign, North Carolina's voter identification law, and the lie spread by Trump and Sen. J.D. Vance that Haitian immigrants were eating pets.
ALSO READ: An indicted lawyer at the RNC’s election integrity unit is still spreading election lies
"Disgusting!" one post reads. "Kamala Harris LIED about the situation in Springfield, OH where Haitians are resorting to eating wildlife and pets in a desperate attempt to survive. @realDonaldTrump was right all along."
This is false. Springfield officials have repeatedly denied any evidence of the claim.
NBC News, the first to report the new study, noted X did not respond to requests for comment but removed many of the posts.
The study found the AI accounts were not supportive of all Republican candidates — only some.
The accounts supported Frank LaRose over Trump-endorsed Bernie Moreno and Arizona congressional hopeful Blake Masters over Abraham Hamadehm, both supported by Trump, NBC News reported.
"Their goal seems to be to shape perceptions in favor of specific conservative candidates, sometimes even at the expense of fellow Republicans," the study states.
"It is possible, if not likely, this network is a domestic campaign. This opens the question as to how many more such networks may yet to be identified."
Clemson University tied the accounts to Dolphin, a large language model system with fewer censorship constraints than ChatGPT. Founder Eric Hartford in a comment to NBC News compared Dolphin to a lighter or a chainsaw.
“I’m producing a tool that can be used for good and for evil,” he reportedly said. He called the deceptive political messaging, "a natural outcome of the existence of this technology, and inevitable."