Logs are telemetry — receipts are enforcement artifacts. The distinction between observation and proof determines whether your AI system is auditable or merely observable.
Logs are telemetry — a record that something happened, written for observation. Receipts are enforcement artifacts — proof that an action was authorized, generated as part of the decision that admitted it. The distinction is whether the system is auditable or merely observable.
An audit artifact written after execution proves observation, not control: it can describe an unauthorized action but could not have prevented it. Runtime governance requires receipts produced inside the enforcement path, before the mutation finalizes, so that the proof and the decision are the same event.