Audit trail

Auditability is not a separate system in Qorpera. Because the wiki is the source of truth, every decision trail — detection, investigation, reasoning, action plan, step results, outcomes — lives on a single page. This makes audit straightforward and makes EU AI Act compliance a structural consequence of the architecture rather than a bolted-on feature.

What’s captured

For every situation:

• Trigger. Which signal, from which source, at what timestamp.
• Context assembly. Which wiki pages were loaded into the reasoning context, with citations.
• Policy evaluation. Which rules applied, pre-reasoning and post-reasoning.
• Reasoning. The four-stage trace — Understand, Investigate, Decide, Produce. Which tools were called. Which reflection invocations fired.
• Action plan. The proposed steps with citations to supporting evidence.
• Human disposition. Approved, edited, or rejected — with the operator’s identity and timestamp.
• Execution results. For each step, the connector response, the outcome, and any errors.
• Resolution. Final status, lessons extracted.

The full lifecycle is readable on one page. No stitching together log aggregators, database records, and stream archives — the situation page is the audit record.

EU AI Act Article 26

Deployer transparency obligations under Article 26 require keeping logs of each high-risk AI system’s operation for a minimum period. Qorpera’s wiki-first architecture means compliance documentation and operational state are the same thing — the situation page is already the log. Admin-tier operators with regulatory requirements can opt into full trace viewing.

GDPR

The audit trail respects data subject rights. Right-to-erasure requests cascade across wiki pages and activity streams; data export produces the full operator-scoped wiki. Audit trails for terminated subjects are retained only as required for legal and regulatory defence.

Governance telemetry

Separate from the per-situation audit trail, aggregate telemetry powers the delegation dashboard: approval rate, consecutive approvals, rejection patterns, correction history per situation type. This is the data operators use to make delegation decisions — data about their own operation, not opinions about what Qorpera thinks they should do.