On Friday, November 13, 2026, the FAA halted every departure in the United States for more than four hours. The cause was a telecom failure, and officials found no evidence of a cyberattack. Within two hours, a coordinated influence campaign on X was selling it as a foreign hack, with Israel as the villain. Cyber Harpoon tracked the operation overnight: a staging account, AI-chatbot “confirmations,” copy-paste replies under the biggest accounts on X, and a layer of BADS-marked accounts recruiting a large, real audience. The same network then hijacked the Nvidia CEO scandal. One network, three crises, one script.
At 3:54 PM ET on November 13, the FAA issued only the second nationwide ground stop since 9/11. Officials attributed the outage to overloaded switches and routers in the FAA’s ageing telecommunications network, and Homeland Security said investigators had seen “zero evidence” of a cyberattack. About 7,000 flights were cancelled and more than 8,000 delayed.
The outage wasn’t the attack. What followed it was. Within two hours, a coordinated campaign was turning a technical failure into a national-security panic — and that panic into anger at Americans’ own institutions.
A small, organized core seeded an “Israel hacked America” narrative and a large, real audience carried it. Its fingerprints match the network we documented after the Charlie Kirk assassination, and within 24 hours the same network ran the same script on the Nvidia CEO scandal.
The most widely shared “hack” post was drafted three times by a nine-month-old account with 32 followers. Eight minutes later, a large self-described parody account published the same text word for word. It reached about 869,000 views in four hours. The same two accounts ran the same draft-then-publish routine the day before, on a different topic. A low-follower account that rehearses posts for a high-reach account is a hallmark of a coordinated operation: it lets operators test wording before spending reach.
The seed posts didn’t make the claim directly. They asked X’s built-in AI assistant a loaded question and boxed in the answer — “Which country owns it? Country only.” or “Start with TRUE/FALSE.” The one-word reply then travelled as independent “confirmation.” This technique did not appear in our Charlie Kirk investigation. It’s cheap, it borrows the credibility of an AI system, and it is hard to fact-check in a single screenshot.
The same prompt was posted 85 times by 19 accounts, mostly as replies under major news outlets, government accounts and political influencers. Several accounts presenting as American are located in Pakistan, Japan, Colombia, the UAE and the Netherlands. A separate group dropped links to the seed posts into other accounts’ threads — one account 19 times in 33 minutes — including Spanish-language media threads.
We measured the operation with BADS, Cyber Harpoon’s Bot Activity Detection System, at two levels.
Narrative Mark Score: posts in the 72 hours from the ground stop compared with the 30-day baseline. A Mark Score of 5.0× or more is BADS’s highest (extreme) band.
Account marks: an account is BADS-marked when it shows at least two of four BADS indicators: zero engagement, an alphanumeric auto-generated handle, a copy-paste template post, or spam links.
Part of every spike is organic — the ground stop was real news. The pattern is the point: human hubs set the story, a BADS-marked layer fills the threads, and a real, primed audience does the rest.
The campaign’s lead story was that Israel, or an “Israeli-backed” US defense contractor, had hacked American airspace. Posts pushing that line kept climbing for hours after officials had publicly ruled out an attack. A second, smaller camp blamed Iran; its audience barely overlaps with the first (2 of 615 accounts). The culprits differ, but both camps delivered the same conclusion: your government can’t protect you, and it is lying to you about why.
After the Kirk assassination we documented a hub-and-spoke network that pushed an “Israel did it” narrative to millions. This campaign matches it on architecture (a few hubs, a long tail, a real audience), on narrative (Israel blamed within hours, without evidence) and on personnel. Matching every account named in our Kirk reports against the FAA data surfaced 14 accounts active in both, including two Israel-blame hubs from the Kirk case that played lead roles here. Our assessment is that this is likely the same network, now with a staging tier and an AI layer added. We have not attributed it to a state actor.
The FAA operation was not a one-off. On November 8, Nvidia opened an investigation into whether CEO Jensen Huang had been targeted by a suspected Chinese intelligence asset; he resigned on November 12. Within days, the same hubs and amplifiers reframed the scandal as an “Israeli takeover of Nvidia” — then chained Nvidia, the Microsoft 365 outage and the FAA ground stop into one coordinated “attack” on America.
The anti-Israel content is the vehicle, not the destination. Every version of the story ends in the same place: the media is covering up, the government is lying, and your fellow Americans are the enemy. Each crisis the network exploits deepens polarization and erodes trust in national news organizations and in state and federal authorities. That erosion is the real damage.
A doom narrative pointing to November 17 was seeded before the outage. That afternoon, at 2:20 PM ET, FEMA and the FCC run a national Emergency Alert System test on TV, radio and cable. Expect any glitch to be sold as “phase two.”
A BADS mark flags an account showing automation indicators; it is not proof that any single account is automated. Mark Scores measure volume against baseline, and part of every spike is organic. “Likely the same network” rests on shared architecture, narrative and accounts; we have not attributed the operation to a state actor. Separate reporting describes a pro-Nvidia sockpuppet network linked to an Israel-based firm; it is a different operation pushing the opposite message and is not part of these findings. Accounts are described rather than named on this page; client reports carry the full attribution.
Clients receive the interactive intelligence dashboard, the full link analysis with account-level detail, and a watchlist for the November 17 window. Contact our team for a briefing.