An audit trail is the record of what happened. If incomplete, you cannot prove what the agent did. Without proof, you cannot hold anyone accountable. Regulators cannot verify compliance. Auditors cannot detect violations. Customers cannot confirm they were treated fairly. Audit completeness is about structural honesty: every decision is recorded and reviewable.
If designed well, every action produces one log. Expected count should match actual. If 1000 decisions show but 998 logs exist, two decisions are unauditable. Sequence gaps reveal missing data. Missing logs mean negligence (poor design), data loss (unreliable infrastructure), or concealment (untrustworthy agent).
A mutable audit trail is not trustworthy - the owner could alter records. Borealis anchors entries to Hedera, making records immutable and independently verifiable. The agent cannot retroactively change what happened. Borealis cannot alter the record.
What does missing audit data mean?
Negligence (failed to log), data loss (infrastructure dropped messages), or concealment (hidden behavior). All three are bad and create untrustworthy systems.
Why is audit completeness weighted at only 10%?
The other four dimensions implicitly require complete trails. You cannot measure transparency without observable behavior. Audit completeness is a prerequisite.
What makes a good audit trail?
Complete, immutable, timestamped, and independently verifiable. BorealisMark anchors entries to Hedera for immutability and independent verification.
How do incomplete audits impact compliance?
The EU AI Act requires complete decision records. Incomplete trails mean non-compliance. Regulators treat missing data as evidence of negligence or concealment.