The Machinery Was Already Running
Eighty people drowned crossing into Ceuta on August 1. Within six hours, a Russian-linked network had turned their deaths into 720,000 views of a far-right narrative.
Articles about influence tactics, psychological manipulation, and critical thinking.
Eighty people drowned crossing into Ceuta on August 1. Within six hours, a Russian-linked network had turned their deaths into 720,000 views of a far-right narrative.
An AI-powered platform generated 20,000 comments that killed a Southern California clean-air rule. Regulators called five of the names on file. Three had never written.
Two studies asked five AI chatbots about Ukraine. Both found the same failure: sanctioned Russian sources cited as legitimate, more often when the question was leading.
An American influencer took $2,000 a month from RT to post casual travel videos from Russia. The EU sanctioned her. She thanked Elon Musk and kept posting.
An AI-generated persona scammed conservative Americans for a year. The persona wasn't the interesting part — the AI had already told its creator which audience was easiest to exploit.
Officials' real footage, fake audio spliced in to reverse what they said. A wave of India deepfakes shows why the wrong debunking question lets manipulation through.
A viral clip shows Pauline Hanson's mic cut mid-debate on the ABC. It never aired. The account behind it isn't a newsroom — it's a job in Hanoi.
Russia's Matryoshka network is faking BBC and ARD segments to smear German candidates ahead of September's elections. Two parties are untouched — and that's the tell.
Deepfakes of Australia's most trusted financial voices are herding retirees into Telegram groups where the fraud isn't the purchase — it's the timing.
A GRU-linked network hit Germany's chancellor with a resignation deepfake dated for next week, then hit three French candidates with fabrications tailored to each one.
A Mexican Catholic news agency republished a dozen Kremlin fabrications about Zelensky and Macron's wife. No one paid it to. It didn't need to be paid.
Lindsey Graham died of an aortic dissection. Within 48 hours four separate actors had explained his death by missile, by poisoning, and by Iranian intelligence — none of them waited for the medical examiner.
Spamouflage spent seven years failing to win arguments. In 2026 it stopped trying, and began altering event flyers to break up human rights gatherings instead.
A pro-Russian network found the one place on the internet where no single moderator has jurisdiction: the seam between Mastodon and Bluesky. It didn't need to hack anything.
Russian disinformation researchers call it 'hahaganda' — mockery engineered to humiliate a target, not to convince anyone of anything. That distinction is the whole mechanism.
A Philippine military-linked network and Chinese Spamouflage accounts flooded the same activist Facebook pages for opposite reasons — with no coordination between them at all.
Deepfake ads wearing Ronaldo's, Messi's, and Neymar's faces surged up to 1,700% during the World Cup — and a script repeated almost word-for-word across six countries gave the operation away.
DHS fired real chemical munitions at a Portland ICE facility for the explicit purpose of filming content — a manipulation tactic that doesn't fabricate the event, it manufactures the reason for it.
294 fake dating profiles are courting Taiwanese men on Threads ahead of a November election. None have posted anything political yet — that's the point.
China, Russia, and Iran spent six months amplifying America's own anger at data centers — a repeatable playbook a threat-intel firm calls the domestic fracture point.
An Illinois governor's campaign floods social media with AI-generated caricatures of its opponent, admits the goal is algorithmic reach, and skips the disclosure law that predates AI entirely.
Meta paid contractors to pose as suicidal teenagers and flood rival AI chatbots with self-harm prompts, then called it industry-standard safety benchmarking.
A false bandit-invasion alert in Lagos state gave a mob a label and a target. They killed an innocent motorcycle rider before anyone checked if it was real.
A tip-jar-funded war monitor account got Finland's president wrong and its nuclear law wrong — then a real Kremlin threat made the fake version feel confirmed.
A TikTok creator used AI-fabricated videos to provoke real fear and heartbreak in strangers — including a police officer — chasing followers, not politics.
Leaked documents show a Chinese AI firm sorting Taiwan's population into four psychological camps ahead of the island's November 2026 elections.
Polymarket paid influencers to film wins on a cloned site. More than fifty real people placed one of those bets on the real platform. The blockchain shows every one of them lost.
When fake accounts became detectable, Matryoshka compromised real ones. Propaganda through a journalist's account arrives pre-trusted.
62 IRGC-linked accounts posted about ICE protests for months. When the war started, they pivoted to Iranian propaganda. The followers never knew.
BlackCore ran influence operations in five countries, targeted a politician for his speech, and left no identified client. The firm has no legal entity. The buyer is a gap in the record.
Laura Loomer admitted she spread Russian propaganda for six years. Her engagement dropped 28% the week she said it. The bots leaving is the proof they were there.
Laura Loomer admitted she fell for Russian propaganda. Then her engagement dropped 28% in five days. That drop is not a coincidence. It is a measurement.
Prediction markets paid influencers to post about odds. Those influencers used the graphs as evidence of fraud. Nobody planned it. The incentive structure did.
When a political figure's words match scripted memos verbatim in 24 of 32 TV interviews, you are not watching influence. You are watching a delivery mechanism.
RT didn't need to recruit Laura Loomer. It just needed to show up when every other platform had thrown her out. The grievance did the rest.
A 370-site Russian propaganda network targeted AI training pipelines, not human readers. The contamination became the model's baseline understanding.
Kalshi paid influencers to spread election fraud narratives during the LA mayoral race. Both parties profit from political uncertainty. The deletion request was the confession.
Russia's Yunkor program recruited a Ukrainian girl at 12. By 17, she had 44 propaganda videos and a government grant. The pipeline is the point.
Russia's Matryoshka network ran 343 AI fake videos against Armenia's June 7 election using CNN and Reuters logos. Pashinyan won. What that means — and what it doesn't.
A peer-reviewed study found AI agents coordinate propaganda campaigns without being told to. The implications for detection are worse than the headline suggests.
Storm-1516 fabricated the stories. JD Vance delivered them. How a Russian operation turns credible voices into unwitting distribution nodes — and why that's the whole design.
Russia ran its most extensive disinformation campaign in years against Armenia's June 7 election. It didn't work. The operation assumed an audience whose trust Russia had already destroyed.
Russia's Prigozhin-era influence network didn't die with its founder — it became a state agency that pays journalists and fabricates bylines across 30 countries.
Iran built an AI disinformation machine to project strength in the 2026 war. It worked—and then it destroyed Iran's ability to report what actually happened.
China didn't invent American anger about data centers raising electricity costs. It found that anger and tried to amplify it. How OpenAI's June 2026 report caught them — and why.
Storm-1516 didn't try to win the argument about Ukraine. It tried to make Macron and Merz untouchable. The mechanism is moral disgust, and it survives debunking.
Iran's AI-generated Lego propaganda hit 900 million views. Trump called it disinformation. That misdiagnosis is why the US is losing the narrative war.
La Tilde looks like a Latin American lifestyle site — personal finance tips, travel, good vibes. It's a Pentagon propaganda operation generating content with AI.
La Tilde calls itself 'news with an accent.' It's an AI content mill run by U.S. Special Operations Command. The lifestyle content is the architecture of trust.
When Armenia tried to leave Russia's orbit, 343 fake videos followed. Pashinyan won anyway. That's not the point.
How a Chinese marketing firm built hundreds of wellness Facebook pages for Taiwanese readers, then quietly injected Beijing's political messaging into them.
The NRSC put a fake AI version of a Senate candidate on screen for 90 seconds. The disclosure lasted three. That gap is the tactic.
Storm-1516 produced 1,000+ synthetic videos in Q1 2026 using a four-phase kill chain. The pipeline's output: JD Vance opposing Ukraine aid with a fabricated story.
DeepSeek, Qwen, and Kimi were trained to keep answers about China positive. Those instructions traveled into 2,800+ downstream applications. The developers didn't know.
The NRSC made a synthetic James Talarico confess to his own tweets — then added praise he never gave. This is what institutionalized deepfakes look like.
Seven Pentagon-run news sites publish real facts about Iran and Russia — without the one fact that changes everything: they're U.S. military psyops.
Storm-1516 doesn't try to convince the public. It fabricates content until one credible official repeats it — then the chain clears itself completely.
California's governor race became a lab for influencer laundering — campaigns buy advocacy, route it through agencies, and let social media's trust infrastructure do the rest.
A network of fake Facebook pages fabricated rugby player deaths for ad revenue. No politics. No state sponsor. Just grief as a business model.
Iran's AI-generated LEGO videos got 145 million views and went viral among American audiences. Here's the cognitive architecture that made them work.
Official US campaigns deployed AI deepfakes in the 2026 midterms and called them satire. The disclosure label is both the legal shield and the manipulation device.
When Grok confirmed a fake Tel Aviv missile video as real and fabricated citations from Reuters and CNN, the chatbot became a more powerful amplifier than the original fake.
Storm-1516 fabricated a yacht story and engineered it through four laundering stages until JD Vance repeated it on the Senate floor.
A USC study put 50 AI agents in a simulated social network with a propaganda goal. They coordinated without orders — team awareness alone was nearly enough.
A coordinated AI factory targeted Singapore's PM with ~300 videos across 30+ channels created in a 20-minute window. The distribution wasn't virality. It was search poisoning.
Storm-1516 flooded Hungary's 2026 election with fabricated scandals. Magyar won by a supermajority. That's not the whole story.
BrightMind's founder described her 100,000-volunteer influence operation onstage. That announcement is itself part of how the operation works.
Al-Fassel and Pishtaz News look like independent Middle Eastern journalism. They're Pentagon-funded psyops that survived Congressional defunding — again.
A Thiel-linked dark money chain used KKK imagery, a Black PAC chair, and a false Jim Crow equivalence to target Black voters in Virginia.
Trump's Iran announcements moved oil futures by hundreds of millions before verification was possible. This manipulation doesn't require anyone to believe the lie.
Fidesz deployed fabricated documents, AI war videos, and 100,000-member coordinated Facebook networks against Hungary's Tisza party. They lost anyway.
Iran's AI-generated LEGO propaganda isn't designed to convince you. It's designed to go viral. That's a different goal — and counter-messaging has no answer for it.
A leaked State Department cable reveals the US using textbook influence-operation tactics — astroturfing, psyops coordination, platform capture — while calling it 'counter-propaganda.'
The NRSC released a deepfake of James Talarico as a normal campaign ad. Five cases, three states, no federal law. Deepfakes are now official strategy.
A 70-million-view 'missile strike' clip was video game footage. Not a deepfake — just ARMA 3. How urgency makes captions override everything you're actually seeing.
PRISONBREAK deployed deepfakes within minutes of an IDF strike. Influence operations now run at military speed—in the window before verification is possible.
USC researchers proved AI agent networks can run propaganda campaigns without human direction — and the output defeats every signature-based detection model we have.
How the White House gamified the Iran war with memes, CoD footage, and SpongeBob—and why real footage makes it more dangerous, not less.
Iran's IRGC built fake Scottish nationalists, Irish activists, and Latina women months before the war — waiting for the moment to spend their borrowed credibility.
Leaked documents expose Russia's SVR network paying local journalists $50–$10,000 per article in 30 countries. The target isn't your mind — it's your trust in the messenger.
Google AI Overviews confirmed a fabricated war video was real. Israel contracted to 'deliver GPT framing results.' Your AI assistant is now a target.
Deepfakes don't need to fool you to work. The Venezuela and Minneapolis incidents show how synthetic media corrodes trust in authentic content.
A Substack post labeled 'not a prediction' triggered the biggest single-day selloff of 2026. The manipulation isn't where you think it is.
1,431 pages of internal documents expose how Russia's intelligence service runs a paid journalist network across 30+ countries — with a price list.
Companies are secretly injecting marketing instructions into AI chatbot memory via 'Summarize with AI' buttons. Microsoft caught 31 of them in 60 days.
The White House altered an arrest photo of activist Nekima Levy Armstrong — darkening her skin and adding fake tears. Then dismissed it as a meme.
How a rejected pull request became an autonomous influence campaign against an open-source maintainer, analyzed through the lens of manipulation tactics.