Global Security, Privacy, & Risk Management

Adversary Amplification in the United States September 6, 2026

In August, X disclosed a network of roughly 200,000 suspected Chinese inauthentic accounts. About 200 of them were posting AI-generated cartoons about American data centers. By then, 70 percent of American voters opposed a data center in their community, New York had a moratorium, and Texas had paused grid connections for every project in the queue. Two hundred accounts did not do that. Tens of millions of Americans did, and almost none of them knew they were part of an operation.

That gap is the subject of the two papers below. Our vocabulary for influence operations describes what adversaries do. Nothing names what happens on our side: the mechanism by which sincere people become the delivery infrastructure for narratives they never knew were seeded. I call it Adversary Amplification. An Agent with intent, foreign or domestic, seeds a narrative onto a grievance that already exists. The commercial platform Ecosystem, built to sell ads, amplifies it because outrage generates engagement. Sincere Nodes carry it forward on their own conviction and credibility. The Impact is never the specific belief. It is distrust, exhaustion, and paralysis, a population that cannot resolve consequential decisions. Once seeded, the machine runs itself and the Agent walks away. It does not even need a foreign hand. The USS Maine, the Iraq WMD case, and the SPLC’s captured hate group designations all ran on the same mechanism. This is the battlefield Qiao and Wang described in Unrestricted Warfare. Their doctrine names the target. Adversary Amplification is how it works.

The companion paper goes one level down to the meme, the compressed image-anchored claim built to be judged in three seconds and shared before the analytical part of your brain shows up. A chart makes a claim feel true even when it proves nothing, agreement feels like evidence, and most people were never taught to ask what else changed at the same time. Fact-checking cannot fix this, because the meme travels without its refutation and organized communities mass-report the correction until it disappears. The durable answer is inoculation: teach the exploitation pattern before people meet the instance, and ask who benefits before you share. You cannot audit everything you see. You can audit what you are about to amplify. A Node that does not share is not infrastructure. The full papers follow.