> ## Documentation Index > Fetch the complete documentation index at: https://docs.elevatedsignals.ai/llms.txt > Use this file to discover all available pages before exploring further. # Audit trail & tamper-evidence > Verify, not just trust — a cryptographically verifiable audit trail. Legacy eQMS audit trails are **append-only relational** records: procedurally secure, but a privileged database user can alter history with no alarm. Invariant adds an **application-layer, per-organization SHA-256 hash chain** over covered audit records, so in-place tampering is **mathematically detectable**. ## How it works (summary) * Each covered audit record is cryptographically linked to the previous one, forming a continuous per-organization chain from an initial anchor — so altering any earlier record breaks the chain. * Every actor type (human, AI, system) and a reason for change are captured. * A **shipped verifier** re-walks the chain and reports any break, identifying the first divergent row. We say **tamper-evident**, not "immutable": append-only rows remain physically mutable by a database owner; our claim is that such tampering is *detectable*, which is the property auditors care about. ## Why it matters for you * A one-click, auditor-facing **chain verification** demonstrates integrity on demand. * The qualification evidence exercises the **same verifier the product runs** — the qualified program is the program you use. * Future-proof: no dependency on any retired ledger service; it is application-layer. ## In the product ### Event Log Covered create, update, approval, e-signature, and status-change events are shown in order with timestamp, actor, action, affected record, and captured details. Regulated electronic records must be attributable and reconstructable (Who, What, When, Why) — this is the single place an auditor, or you, can trace any decision end to end. During an investigation or audit prep, locate the affected record here and walk its history chronologically before drawing conclusions. ### Actor Types: Human, AI, System Each entry is badged by who performed it — a named user, the AI quality layer, or an automated system process. Human entries show the user's email. As AI participates in quality workflows, attribution can't blur; separating human, AI, and system actions keeps accountability explicit and reviewable. ### Verify Integrity One click recomputes the stored chain and checks each covered row's hashed content and linkage, reporting the exact row when a mismatch is found. A passing result demonstrates that the stored chain is internally consistent at that as-of time — it does not prove completeness against a privileged wholesale rewrite or tail deletion without an external anchor. Run verification during data-integrity review and retain the as-of result with other evidence; do not present a green result as a full validation or completeness conclusion. ### Change Details Each entry records the specific fields that changed, so you can see the substance of an edit — not just that one occurred. When reviewing a deviation or CAPA history, compare successive entries to confirm edits match the documented rationale. Technical statement + qualification evidence: see the Validation Accelerator Pack (artifact A6).