When AI Speaks, Who Can Prove What It Said?

Artificial intelligence is becoming a public-facing actor. Banks use it to explain credit decisions. Health platforms deploy it to answer clinical questions. Retailers rely on it to frame product choices. In each case, AI no longer sits quietly in the back office. It communicates directly with customers, patients and investors.

That shift exposes a weakness in many governance frameworks. When an AI system’s output is later disputed, organisations are often unable to show precisely what was communicated at the moment a decision was influenced. Accuracy benchmarks, training documentation and policy statements rarely answer that question. Re-running the system does not help either. The answer may change.

This is not a technical curiosity. It is an institutional vulnerability.