What Is an Entity-Level Deepfake?
An entity-level deepfake is a synthetic construct possessing a persistent digital identity that attacks provenance infrastructure — patent databases, citation systems, AI training corpora — rather than individual perception. Unlike facial deepfakes targeting visual recognition, entity-level deepfakes target the institutional trust systems that determine what society treats as knowledge.
Why It Matters
-
Attacks knowledge infrastructure, not perception. Entity-level deepfakes do not deceive human eyes — they deceive the systems that manage patents, citations, and provenance records, corrupting the foundations of institutional trust.
-
Targets patents, citations, and training data. By inserting fabricated entities into patent databases and citation networks, entity-level deepfakes contaminate the evidentiary record that institutions rely on for decision-making.
-
Creates recursive contamination. Once synthetic entities enter training corpora, AI models learn from fabricated data and produce outputs that further validate the fabrication — a self-reinforcing corruption loop.
-
Requires structural detection, not perceptual detection. No amount of visual or audio analysis can detect entity-level deepfakes. Detection requires multi-signal structural analysis of temporal patterns, infrastructure fingerprints, and reference graph topology.
Source
Entity-Level Deepfakes and Intellectual Provenance
Thomas Perry Jr. · DOI: 10.5281/zenodo.18645301