The AI-Driven Patterns of the Index

Most manipulation techniques are static — a fixed trick, shipped once. A growing class is not. This section gathers every record in the Technique Index flagged AI-driven — from the relatively benign ML-timed re-engagement and engagement-optimized ranking, which merely learn your habits, to the covert and classified constructs the field teams have logged but never fully verified: Emotion Surveillance, Generative Persuasion, the Sentinel Compulsion Stack.

THE RESET tracks these separately because a self-optimizing pattern is a different kind of threat: it adapts faster than a survey can fix it, personalizes past the point any auditor can reproduce, and leaks almost nothing a static analysis would catch. The Instrumentation Mesh runs behavioural and inference-trace sensing across the surfaces so the adaptive layer is logged as carefully as the fixed one. The classified constructs matter most here: their behaviour is often the only trace they leave, and chasing that behaviour is how the Directorate tells a real deployment from a model’s hallucination.