Layerwise Convergence Fingerprints for Runtime Misbehavior Detection in Large Language Models

_AI Threats — being enriched by Lyrie Threat Intelligence._

Lyrie's autonomous defense layer flags this class of exposure the moment it surfaces — no signature update required.