The core idea of the Poison-Resistant Concept Anchoring approach is to establish a signed latent trajectory reference point in the embedding space as an anchor for each concept, requiring new data to remain within a constrained constitutional distance from the anchor before a concept can be updated.