Hundreds of apparent pro-Russian bot accounts on X are pushing US election misinformation and amplifying false narratives about Democratic contender Kamala Harris, a research group said Thursday, calling them "sleeper agents" for having evaded detection for years.
The findings by the Washington-based American Sunlight Project (ASP) demonstrate how bot-like activity plagues X, previously called Twitter, despite pledges by billionaire owner Elon Musk to crack down on the digital manipulation.
ASP analyzed nearly 1,200 accounts, a long-standing network that generated more than 100 million posts as of July, including pro-Kremlin propaganda, content favoring Republican nominee Donald Trump, and misinformation about Harris's campaign.
The accounts, some of which have escaped detection and moderation on the site for as long as 15 years, retweeted such content within seconds of its posting, indicating bot activity, the group said in a report shared with AFP ahead of its public release.
"We were not surprised to find another pro-Russian bot network, but we were shocked to learn that some of the accounts in the sleeper agent network have been active for more than a decade," Nina Jankowicz, the group's co-founder and chief executive, told AFP.
Jankowicz, the former Department of Homeland Security disinformation chief, called on X to take down the network, which has seen an uptick in "abusive and false content" targeting Harris.
One account created in 2020 promoted the falsehood that Harris had admitted that she will be a "puppet" of Ukraine's President Volodymyr Zelensky if elected president.
It also touted the unfounded claim that the White House was pushing for regime change in Lebanon, taking advantage of Israel's recent attacks on the militant group Hezbollah.
Another account created in 2011 shared a post by Musk -- who has endorsed Trump and courted criticism for amplifying political falsehoods through his influential personal account -- that pushed the debunked narrative that migrants were being imported into the United States to manipulate the November 5 election.
Hundreds of accounts in the network are not attributable to real social media users, with some creating fake personas using images from stock photo websites such as Shutterstock, ASP said.
To disguise their objectives and more easily "inject themselves into larger X/Twitter conversations," some accounts regularly shared content about subjects such as sports and cryptocurrency, the report said.
It was not possible to determine the precise entity behind the pro-Russian accounts.
With data restrictions imposed by X since Musk purchased the company in 2022 for $44 billion, it was also difficult to assess their exact reach.
Researchers are now required to pay a hefty fee for access to its API, which allows third-party developers to gather the social platform's data.
"If researchers had data access restored, more of such activity would likely be visible," the ASP report said.
Bots and other automated accounts, researchers say, are a cornerstone of the Kremlin's efforts to spread misinformation, in some cases supplanting state media accounts which have been restricted across several countries since Russia's full-scale invasion of Ukraine in February 2022.
X did not reply to AFP's request for comment.
Ahead of his purchase of the platform, Musk pledged to "defeat the spam bots or die trying."
But bot activity remains entrenched on the platform, a report from Australia's Queensland University of Technology said last year, after an analysis of about one million posts.
The platform has gutted trust and safety teams and scaled back content moderation efforts, making it what researchers call a hotbed for misinformation.
"Despite the fact that Musk has an avowed goal of ridding his platform of bots, we've found that they persist on X, even coming from networks that are likely state-affiliated," said Jankowicz.
"This is behavior that is fairly easy to identify, and yet this multi-billion dollar corporation has not cracked down on these accounts that violate its platform manipulation and spam policies."
