Hundreds of fake accounts defended Jensen Huang and attacked lawmakers over chip exports.

Late last week, the Financial Times reported that Jensen Huang, the CEO of Nvidia, was believed by U.S. intelligence to have had an affair with a Chinese spy. The alleged spy was a business executive whom American officials warned China recruited to gather information on the company and potentially push for chip exports to China.

In the following days, as the story bounced around social media, dozens of accounts rushed to Huang’s defense. They dismissed the report as resting on anonymous sources and insufficient evidence. Six of them said, in nearly the same words, that Huang had already made clear in May that Nvidia’s Blackwell and Rubin chips weren’t going to China — and thus concerns about the chip export angle were overblown.

The accounts belong to a network of over 400 apparent sockpuppet accounts that appear to have spent years promoting messages for various clients, including, over the past year, taking Nvidia’s side in Washington’s fight over selling advanced AI chips to China. The same network is actively supporting other causes, including backing the Rapid Support Forces in Sudan’s civil war, a group the United States has accused of genocide.

In a comment to The Midas Project, an Nvidia spokesperson said the company recently became aware of these accounts and is trying to identify the source but has so far been unsuccessful.

Since November 2025, when the network became active on this topic, the accounts have posted thousands of times, generally taking Nvidia’s side on hot-button discussions across X, Bluesky, and Reddit. Their primary topic has been export controls, the rules limiting which AI chips U.S. companies can sell to China. In addition, they’ve defended Nvidia as a company and Huang personally, dismissed the severity of chip smuggling, attacked Anthropic for its views on AI safety, and generally written off concerns about the technology and calls to slow the industry down.

The accounts also single out the people pressing hardest for limits on chip sales. No one draws their attention more than Chris McGuire, a former National Security Council official who is now a senior fellow at the Council on Foreign Relations and a leading advocate of export controls. Dozens of accounts in the network have replied to him more than 100 times since November 2025. “You’re incredibly loud but no substance!” one told him in August. Other frequent targets include Sen. Elizabeth Warren (whom one account calls “pure evil”), the House select committee on China, the Republicans on the House Foreign Affairs Committee, Michael Sobolik (a China specialist at the Hudson Institute), and Samuel Hammond (chief economist at the Foundation for American Innovation).

Their replies to lawmakers who criticized chip sales to China often turned to insults: they replied “Peabrains” to a post by the House China committee and its chairman, Rep. John Moolenaar; told Rep. Bill Huizenga “U r really making me question your competence sir”; dismissed Sen. Chris Van Hollen’s criticism of Nvidia’s chip sales to China as “fear bait” and “political bs”; wrote off Sen. Chris Murphy’s similar warnings about Nvidia as “BS”; and called a post by Rep. Sydney Kamlager-Dove “anti-Trump propaganda” and “simply a lie” (a reply that account posted three separate times in eight minutes).

Various posts from the network also attack Dwarkesh Patel (who had a heated debate with Huang that went viral) and Dario Amodei (who has taken the opposite position from Nvidia on chips and safety concerns).

Of the more than 400 accounts The Midas Project has identified, most present themselves as Americans: X lists about 80% as based in the United States, though for nearly all of those it warns that the location “may not be accurate,” suggesting the account connects through a VPN or proxy. They pose as everyday users, including day traders, a nurse’s aide, a data scientist, a pencil-sketch artist, and environmentalists. Between campaign posts, they chat about sports, pets, and celebrities.

Several lines of evidence tie these accounts together. Most were set up in batches, created within an hour of another account in the same network, across dozens of individual account-creation sessions. The largest session lasted two and a half hours on Nov. 9, 2021, during which about 30 accounts were created. The vast majority of accounts in the network were created, renamed, or given new profile pictures or banners in the same sessions as other accounts in the network, often minutes apart. Many also seem to release their posts through scheduling software, with the post going live within the first half-second of a new minute. At times, the accounts will also slip up and post virtually identical text on two or more different accounts.

Nvidia isn’t the only shared interest that these accounts have. Over the years, this network of accounts has advocated for a variety of causes, seemingly on behalf of both corporate and political clients. This has included campaigns benefiting Israel, pharmacy benefit managers, manufacturing companies, and, most recently, the Rapid Support Forces in Sudan. One of the most distinctive traits of accounts in this network, beyond being set up or reskinned in the same session as other accounts, is their participation in multiple highly specific campaigns.

The network’s best-documented campaign, and its largest by volume, appears to be one exposed by the Israeli newspaper Haaretz: “Project Lighthouse,” a pro-Qatar effort that Haaretz reported was devised by a firm run by two aides to Prime Minister Benjamin Netanyahu. In its reporting, Haaretz named 15 of the campaign’s fake accounts. Most of those named accounts have since participated in the pro-Nvidia campaign, and dozens more are connected to these Haaretz-named accounts through shared account setup and reskinning sessions.

Haaretz reported that Koios, a small Israeli company that presents itself as a firm that detects financial fraud online, ran the fake accounts. Koios denied any involvement. Israeli police have since raided its offices as part of their investigation into Qatar’s ties to Netanyahu’s office, and as of September 2025, the company claims to have ceased operations. But the social media accounts did not stop. Soon after, they began their campaign on AI chips and Nvidia.

Indeed, Koios’ business appears to have carried on under a new name: IntellACT (sometimes written as IntellACTS). Not to be confused with the identically named Israeli aviation company (a choice that may have been designed to make the company harder to trace), IntellACT seems to have emerged around the time its website was registered in September 2025, the day before Koios’ liquidators were appointed, and the same month Koios said it ceased operations.

IntellACT’s website was registered Sept. 8, 2025. The very next day, liquidators were appointed for Koios.

Web searches for IntellACT turn up little, but they do reveal that an Argentina-based individual who was once Koios’ “Web and Financial Intelligence Team Leader” has two résumés online describing the same job with the same start date, one at Koios and the other at IntellACT. The same person has posted IntellACT job ads for Argentina-based employees. And on Oct. 6, 2025, weeks after Koios said it ceased operations, that person advertised for a “Social Media Manager for Public Companies” to manage X and Reddit accounts for “an international company with US clients”; a LinkedIn user who lists Koios as their employer since 2024 reposted the ad.

Two online resumes for an employee who has posted jobs for IntellACT show the same start date and similar duties, but different employer names: Koios and IntellACT.

Meanwhile, Koios’ sales chief, whose hiring Koios announced in 2024, now describes his role on X as “CRO @ IntellACT,” with a link to IntellACT’s website. His LinkedIn profile still lists him as Koios’ chief revenue officer.

Koios’ Chief Revenue Officer now holds the same role at IntellACT.

Both companies sell themselves as defenders against online manipulation. Koios offered companies alerts to “negative campaigns so you can counteract,” along with help to “get your narrative out” through “top influencers.” IntellACT advertises “uncovering negative campaigns” and promises to “neutralize coordinated attacks.” It’s not clear whether IntellACT’s website is actively used in the course of its business — among other things, it randomly generates statistics in the “operational snapshot” section using a script, and its contact form sends mail to a template placeholder address. 

But IntellACT seems to be active nonetheless: recent job ads from the company seek writers for clients’ “investor relations channels and responder programs” and a Twitter/Reddit account manager in Argentina or Vietnam for $500 to $800 a month. Strengthening the link to Koios, Haaretz’s sources had said Koios employed staff in Argentina.

An IntellACT job ad looking for someone to run Twitter and Reddit accounts out of Argentina or Vietnam.

Between the job listing for social media account operators, the posting habits of these accounts, and even occasional operator mistakes discernible from their posts, it appears these are not bots but sockpuppets. That is, the accounts are likely human-operated, with workers scrolling social media manually and typing replies to posts, often with the help of AI and likely a script. In fact, the most revealing evidence of this comes from one operator mistake where a network account, presumably trying to create lifestyle filler content, wrote, “Create a reply on X (Twitter) for this post. The goal is to get a lot of engagement while still sounding like a real human. Use emojis if appropriate. 280 characters or less. Provide 3 options.” Four different posts also call Nvidia “NCore,” which may be a codename from a brief the operators work from.

Comparing the accounts against the web archive also reveals that deleting posts is common for accounts in the network. This tracks with what sources told Haaretz in 2025: that clearing out old posts was standard practice for Koios. The accounts regularly remove their campaign posts weeks or months after publishing them, while leaving their everyday chatter up. Of the over 4,000 posts about chips and AI the network has made since November 2025, approximately two-thirds are already gone. And in the hours before this article was published, dozens of the network’s accounts were suddenly deactivated, and others began deleting posts supporting Sudan’s Rapid Support Forces.

I have previously covered two separate social media campaigns focused on chip and data center policy that I speculated Nvidia may have been behind. I also covered a similar campaign promoting videos from Innovation Council Action, a pro-industry group incorporated around the time this Nvidia campaign began (and which I similarly speculated could have links to Nvidia). But when asked for comment on this campaign, Innovation Council Action suggested it wasn't aware of it. IntellACT didn’t respond to my attempts to reach it, while the Koios liquidator claimed that no company assets were sold or transferred and that there is no record of interaction between Koios and Nvidia. Koios said via a generic company email that it “has no connection to this matter” and that it ceased operations in September 2025. And an Nvidia spokesperson said in a statement: “We have no relationship with those firms and have not asked any other party to engage them on our behalf. Recently, we also became aware of these accounts and are attempting to identify the source, which has so far been unsuccessful.” With all this in mind, who might have been funding a complex, yearlong operation like this remains unclear and may never be known without a more formal investigation.

Could a formal investigation be warranted? I certainly think so. This appears to have been a yearlong campaign designed to shape one of the most important policy debates in Washington. A network of hundreds of fake accounts made thousands of posts arguing Nvidia’s side of the debate over selling AI chips to China, defended its chief executive through a spy scandal, and went after the commentators pushing hardest for export controls, including U.S. lawmakers — only to later delete much of what they said. These accounts are directly linked to a supposedly defunct Israeli influence firm that police raided over a covert pro-Qatar campaign, and the firm has since filed for liquidation. The same accounts are currently running other campaigns, including one backing Sudan’s Rapid Support Forces, a group the United States has accused of genocide. 

Who is ultimately behind this campaign, we don’t know. But the chips at the center of this fight may be the most strategically important technology of our time, and whether China gets them is being decided now. A debate this important can be quietly influenced by manufactured voices. It can be done cheaply, apparently from abroad, with the evidence scrubbed afterward. That should worry anyone who cares about the policy decisions that result, no matter which side they’re on.