Social media platform X has suspended at least 15 accounts suspected of spreading Russian disinformation about the 2026 United States midterm elections, just days before voters head to the polls on Tuesday, November 3. The suspensions, which reportedly took place on Thursday, October 29, included some accounts with more than 100,000 followers, according to the disinformation watchdog the Institute for Strategic Dialogue (ISD), as reported by CNN. Researchers say the action shows X catching some covert Russian-linked activity quickly, but warn that other influential accounts and parts of the same network remain online.
With early voting already under way across the country and more than 43 million ballots cast before the final weekend, the suspensions land at the most sensitive moment of the campaign. Experts say Russian interference in the 2026 midterms has been relatively limited so far, but they have recently detected a rise in activity from suspected Russian-linked accounts spreading false claims about several Democratic Senate candidates. The latest enforcement is therefore being read as both a sign that platforms are watching and a reminder of how fast misleading posts can travel in the final days before an election.
According to the ISD, X took action on Thursday against at least 15 accounts identified as part of suspected Russian disinformation activity aimed at the midterm elections. The watchdog said several influential accounts promoting similar claims were still active, even after the suspensions. CNN reported that it had contacted X, formerly known as Twitter, seeking comment on the account suspensions.
The suspensions were the latest step in a fast-moving final week. Earlier in October, X announced a much larger sweep, saying it had removed more than 13,000 accounts it suspected were part of three foreign influence operations linked to Iran, Russia and China. That disclosure, made on October 14, said the accounts used inauthentic identities to post and amplify political content about candidates and public figures across the political spectrum. X did not give a country-by-country breakdown of the 13,000 accounts, and it has not published technical evidence for its country attributions.
The new, smaller action is different in an important way: it is tied by researchers to a specific, recent wave of misleading posts about Democratic Senate candidates, rather than to a general sweep. ISD researchers say the accounts were part of activity spreading false claims designed to damage those candidates in the closing days of races that could help decide control of the Senate. The Senate currently stands at 53-47 in Republican favour, and Democrats need a net gain of four seats to take the majority.
The suspensions follow weeks of scrutiny of AI-linked videos targeting two Democratic Senate candidates. One video targeted Sherrod Brown, who is running against Republican Senator Jon Husted in Ohio. Styled to look like a newscast, it used photographs of Haya Akbik, who interned for Brown several years ago, and other former staff members, and attributed claims to her that she has firmly denied. Akbik told The Associated Press that any claims attributed to her in the clip are "completely made up", adding: "It's my name, it's my reputation, making this allegation, and it's just completely false."
A second video targeted James Talarico, the Democrat running against Republican Attorney General Ken Paxton for a Senate seat in Texas. Framed as a direct-to-camera confession from a former student of Talarico, it was posted to X on October 20 by an account created the previous month. Fact-checkers found that the person in the video was not the real former pupil, who has denied the claims, and PolitiFact rated the underlying allegation false. A spokesperson for Talarico said opponents were resorting to fake videos and conspiracy theories because they could not defeat the candidate on the truth.
Media forensics experts link both videos to Storm-1516, a Russian network tracked by researchers since the 2024 presidential campaign. Clemson University professor Darren Linvill said the new videos bear similar hallmarks to Storm-1516 material that targeted former Democratic presidential candidate Kamala Harris and her running mate, Minnesota Governor Tim Walz, in 2024. Dartmouth College professor Hany Farid, who studies AI deepfakes, assessed that the Brown video was a "fairly low-tech fake" while the Talarico video appeared to be AI-generated. X removed the initial posts sharing the videos and suspended the users who posted them, but copies continued to circulate, some with hundreds of shares.
The Russian Embassy in Washington did not immediately respond to requests for comment on the videos, and has in the past denied interfering in United States elections. The US Cybersecurity and Infrastructure Security Agency, which has publicly identified some videos as Russian disinformation in past election cycles, also did not respond to requests for comment.
Bret Schafer, senior director at the ISD, said X had taken action against some of the accounts identified as part of the suspected disinformation activity, but warned that the job was only partly done. "X has shown that they have a beat on this, to some degree," Schafer said. "A lot of the covert Russian posts that we've seen go up have been taken down quickly. But other parts of the infrastructure have been left up."
That distinction matters, researchers say, because influence operations rarely depend on a single account. A typical playbook, described by Linvill and other researchers, starts with a false claim seeded on social media or on a fake website posing as a news organisation or think tank. The claim is then laundered through influencer accounts with large followings, so that it reaches real voters who may never see the original post. Removing the covert accounts helps, but leaving influential amplifiers online means the same claims can keep spreading.
Experts also caution against overstating what a suspension count proves. Fifteen suspended accounts, or even 13,000, do not by themselves show how many voters saw the content or whether it changed anyone's vote. What the suspensions do establish is that a platform found accounts it believed were breaking its rules against coordinated inauthentic behaviour and took them down. Whether the operation changed the campaign is a separate question, and one that researchers say cannot be answered from takedown numbers alone.
Schafer's assessment was measured rather than celebratory: X is catching covert posts quickly, but the wider infrastructure survives. That view echoes the experience of the people targeted by the earlier videos. Akbik's denial was blunt and personal, focused on the harm of having her name and photograph used to make claims she says are entirely false. Talarico's campaign framed the videos as an attempt to smear a candidate through fabrication rather than debate.
Platform accountability researchers have also pointed to the limits of X's transparency. In its October 14 announcement, the company said its Safety team identified the 13,000 accounts through proactive sweeps and investigations under its rules on spam and coordinated inauthentic behaviour. But it did not name the candidates targeted, publish the narratives involved, or release technical indicators supporting its country attributions. Independent reporting likewise found no public technical breakdown of those operations. The same gap applies, for now, to the latest 15 suspensions: the accounts have not been publicly named, and X has not commented on the action.
The timing adds to the scrutiny. With Election Day on November 3, there is little time for corrections to catch up with a viral falsehood. Researchers note that the 2024 precedent is instructive: US intelligence agencies assessed that Russian government operatives were behind several fake videos and other misleading online content intended to influence that election. The 2026 activity, they say, bears similarities to those 2024 operations, including the use of staged videos and fake personas.
Foreign influence operations do not need to hack a voting machine to matter. Their goal, researchers say, is usually cheaper and subtler: to inject noise, deepen distrust and suppress or sway a slice of voters in close races. That is why Senate contests in Ohio and Texas, both rated competitive in recent polling, are such attractive targets. In Texas, polls have shown Talarico narrowly ahead of Paxton or the race essentially tied; in Ohio, Brown and Husted have traded small leads. In races that close, even a modest shift in enthusiasm or turnout could matter.
The suspensions also matter for what they reveal about the final days of a campaign. Voters are now deciding with incomplete information, often on their phones, and a fabricated video can travel further in hours than a fact-check can in days. The Brown and Talarico cases showed that pattern clearly: the original posts were removed, yet copies persisted, and in one case AI assistants repeated the false Talarico claim, citing articles that never mentioned it. Platforms, researchers and newsrooms are therefore under unusual pressure to act fast and to show their work.
There is a wider political context too. The midterms arrive with President Donald Trump at a new low in Emerson College polling, at 37 percent approval, and Democrats leading the generic congressional ballot 52 percent to 42 percent in the same survey. Republicans are defending narrow majorities in both chambers, and prediction markets favour Democratic majorities. In that environment, every new claim about interference, and every platform action against it, becomes part of the campaign argument itself.
Early voting ends on Saturday, October 31, in most counties, and Election Day follows on Tuesday, November 3, when all 435 House seats and 35 Senate seats are decided. Expect platforms to face fresh questions in the coming days about any new suspicious accounts, and expect researchers to publish further findings as they trace the networks behind the suspended accounts. X may also be pressed to say whether more accounts linked to the same operation remain active, and to explain what evidence supports the Russian attribution.
Voters, meanwhile, have a practical defence that does not depend on any platform: treat dramatic, anonymously sourced videos about candidates with scepticism, check whether established news organisations have verified the claim, and be wary of posts that use a real person's name or photograph to make an allegation that person has denied. Election officials have also stressed that the administration of voting itself, which is run by states, is separate from the online information environment, and that misleading posts do not change ballots already cast.
If further suspensions follow before Tuesday, they will likely be reported in the same piecemeal way: a watchdog finding, a platform action, and a debate about what was left online. The ISD's warning is the thread to watch. Takedowns of covert posts are happening quickly, Schafer said, but the infrastructure behind them has not been fully removed.
At least 15 accounts, according to the Institute for Strategic Dialogue, with the suspensions reportedly taking place on Thursday, October 29. Some of the accounts had more than 100,000 followers.
Spreading suspected Russian disinformation about the 2026 midterm elections, including false claims about several Democratic Senate candidates. The attribution to Russia is the researchers' assessment of suspected activity; X has not publicly detailed its own evidence for these specific accounts.
Researchers link the wider wave of activity to AI-linked videos targeting Sherrod Brown in Ohio and James Talarico in Texas, which experts attribute to the Storm-1516 network. X previously removed the initial posts sharing those videos and suspended the users who posted them, but copies kept circulating.
CNN reported that it had contacted X seeking comment. No X response was included in the initial reporting of the suspensions.
Election Day is Tuesday, November 3, 2026. Early voting ends on Saturday, October 31, in most counties, and more than 43 million ballots had already been cast before the final weekend.
Related coverage: Russian AI deepfakes target Brown and Talarico in final days before midterms, Trump approval rating hits new low at 37% in Emerson poll before midterms, and Ted Cruz warns Republicans could lose every statewide race in Texas.
Leave a Comment
| Today's Date:
