Signal provenance is where a claim came from, made inspectable. In an AI society it matters enormously to be able to ask of consequential information: was this directly observed, measured, inferred, generated by a model, supplied by an interested party, independently verified, or translated through several intermediaries? The model treats provenance almost as infrastructure — not because everyone needs to inspect every chain, but because when something matters, the chain should be available to inspect, which is what makes manipulation harder.
Provenance over suppression
Provenance is also the model’s preferred alternative to letting a central authority define truth. Rather than suppressing a claim judged false, the healthier response is usually to expose provenance, evidence, uncertainty, competing claims, and consequences — keeping signal alive without making all claims equivalent, and distinguishing error, deception, uncertainty, dissent, and emerging signal instead of collapsing them together. It pairs with the requirement that important public signal stay layered (raw source through accessible summary) so the raw layer remains reachable and AI does not become an epistemic gatekeeper. See How Signal Should Flow.