Deepfake student smears Texas Senate candidate

TLDR

• Graphika assesses with high confidence that Russia-aligned Storm-1516 produced two deepfake videos accusing Texas Senate candidate James Talarico of assault.¹
• The man whose identity was copied told Lead Stories the video is not him.²
• NPR, citing Alethea, reports one video drew about 857,000 views in 24 hours; X later suspended the originating account.⁴

Researchers at Graphika, Alethea and Clemson University attribute two AI-generated videos posted on 20 October, in which a man claims Texas Democrat James Talarico sexually assaulted him as a schoolboy, to the Russian operation known as Storm-1516.¹⁴⁵

Why it matters: the case shows the checks that separated this fake from a genuine accusation. The posting account was new and impersonated a real person, the real person denied it when contacted, and the clips carried visible deepfake tells.¹² Researchers also traced the spread through accounts they already associate with the operation, which is a repeatable method for attribution.¹⁴

Driving the news:

• Graphika says an X account created in September 2026 posted the two videos on 20 October under the name of a real Facebook user from Texas who attended the middle school where Talarico taught. The videos show a man claiming Talarico assaulted him in January 2012, when he was 12.¹
• Graphika cites unnatural face movement out of sync with the audio, little body movement and AI-detection tools. Lead Stories ran both clips through the Hive detector, which scored them 99.1 and 99.8 per cent likely AI-generated, and found the voice synthetic. The real man told Lead Stories the video was "very bad AI" and that Talarico never abused him.¹²
• NPR reports a network of 15 accounts that Clemson's Darren Linvill flagged as bots shared the post at once. About 12 hours later three large X accounts that researchers say previously amplified Storm-1516 narratives shared it, and one has solicited paid partnerships.⁴
• One of those amplifiers, SGAnon, shared the video and deleted it after saying its own analysis found it "artificially generated", according to Graphika.¹
• PolitiFact cites NewsGuard as reporting that X gave Romania as the account's signup country. NPR says the profile claimed a US location but X flagged a possible proxy or VPN.³⁴
• By late on 21 October X had suspended the originating account and one large amplifier, and two others had removed their posts, NPR reports.⁴ PolitiFact rated the claim "Pants on Fire", and a Talarico spokesperson said opponents were slandering him with AI deepfakes and foreign bots.³

The big picture: Linvill says Storm-1516 often swaps AI faces onto real actors and, in 2024, made a similar baseless claim against Tim Walz using a paid actor from West Africa.³⁴ He argues the aim is to entrench views among people already hostile to the target rather than to persuade Talarico's supporters.³ NBC News calls it the most prominent example yet of a foreign effort to mislead voters before the midterms.⁵

The other side: attribution rests on researchers' assessments of tactics and amplifier networks, not on an official finding. Representative Jim Himes told reporters nothing suggests Russia can or intends to interfere with voting mechanics.⁴ View counts differ between outlets: Meduza, relaying NBC, says one post passed 1.3 million views.⁶

Single-source note: the amplifier names, the 15-bot cluster and the 857,000 figure come from Graphika, Clemson and Alethea via NPR and have not been independently verified here. Detector scores are probabilistic. The NBC News article was readable only in part.

Update, 27 October 2026: NewsGuard reports that AI tools repeated the fake accusation as fact. When it searched the names on 21 October, Google's AI Overview said the supposed former student "published a video claiming" Talarico assaulted him, citing ABC News and KTAL articles that never mentioned the allegation, and Google's Gemini and Microsoft's Copilot gave the same false account.⁷ Google and Microsoft did not respond to NewsGuard's emails.⁷ NewsGuard also notes the man in the videos has teeth that differ from the real person's in authentic photos, speaks with an accent unconnected to the real person's record, and used the British spelling "behaviour" in a caption, and that the Gnida Project, an anonymous research group, attributed the videos to Storm-1516 on 21 October.⁷ NewsGuard puts the first post at 33,000 views and 200 likes after one day, well below the 857,000 reported by NPR and cited above, so view counts remain inconsistent between sources.⁷ In a policy brief on 26 October, the Foundation for Defense of Democracies argues that platforms should suspend repeat Storm-1516 amplifiers, that AI products should ensure cited articles contain what the answer claims, and that Washington should name foreign operations quickly, noting the US government had not publicly attributed this one.⁸ Those recommendations are the think tank's position, not findings.⁸

Update, 28 October 2026: NewsGuard reports that the same influence operation has since targeted Sherrod Brown, the Democratic Senate candidate in Ohio, with a fabricated television news segment.⁹ A pro-Russian X account posted the video on 26 October, styled as a Cleveland 19 broadcast saying four former staffers had accused Brown of sexual assault, and NewsGuard says it added dubbed audio to authentic footage from a 21 October segment in which the anchor discusses an attack ad against Ohio governor candidate Amy Acton.⁹ The station's news director told NewsGuard "We have reported no such thing", and one of the four people named in the video, a former Brown intern, called the claim false.⁹ X had suspended the account by 27 October, but a copy shared by another account reached 536,700 views by the next morning with no Community Note, according to NewsGuard.⁹ NewsGuard bases its Storm-1516 attribution on the video's format and amplification pattern, and says the Gnida Project, an anonymous research group, attributed it to the operation on 26 October.⁹ The Associated Press reports that the social media videos, believed to be produced with AI and linked to Russia, misuse real people's names and photographs to spread disinformation about two Democratic Senate candidates.¹⁰ Only the opening of the AP article could be retrieved for this update, so its further detail is not reflected here.¹⁰

Update, 29 October 2026: Lead Stories fact-checked the fake Cleveland 19 segment about Sherrod Brown and found no matching report on the station's website or in Google News and Yahoo News searches.¹¹ It ran the first nine seconds of the clip through the Hive Moderation detector, which rated the speech 88.4 per cent likely to be AI-generated, an estimate from a detection tool and not proof.¹¹ Lead Stories also compared the fake with the station's real 21 October broadcast: the first words match, then diverge, and both show the same time and temperature on screen.¹¹ Using extracted timestamps, it traced the earliest post it found to 6:58 p.m. UTC on 26 October, more than seven and a half hours before the account it had first examined, and says X has since suspended that earlier account.¹¹ Straight Arrow News reports that disinformation researchers attribute the Brown and Talarico posts to Storm-1516, and that Darren Linvill, founder of Clemson University's Media Forensics Hub, says such operations reuse existing narratives to make Western democracies appear corrupt.¹² The outlet adds that researchers tied similar unsubstantiated sexual abuse claims against Tim Walz in 2024 to the same group, and that the accounts behind the first Brown and Talarico posts have been removed by X.¹²

Update, 30 October 2026: NewsGuard reports that a third Democratic Senate candidate, Maine's Troy Jackson, has been targeted with a fabricated TV segment that it links to Storm-1516.¹³ The video, posted on X on 28 October by an account that says it is based in Belarus, imitates a WABI (Bangor) news segment and claims staffers say Jackson threatened to cut them in half with a logging axe.¹³ NewsGuard says the clip is dubbed audio over real WABI footage from 21 July, which concerned an alleged water-bottle incident, and WABI's vice president told NewsGuard the video is fabricated and appears to use synthetic voice.¹³ The original post drew 4,400 views, but later shares reached hundreds of thousands, with no Community Note, per NewsGuard.¹³ Gnida Project attributed the video to Storm-1516 on 28 October, and NewsGuard says its format and amplification pattern suggest the same operation, which is an assessment and not a confirmed attribution.¹³ The Associated Press reported on 27 October that videos believed to be AI-made and linked to Russia are misusing real people's names and photographs to spread disinformation about two Democratic Senate candidates.¹⁴ The AP piece was not read in full at the time of writing.