Compliance Report

Regulated agentic ABM run · generated 2026-05-30T11:09:41.364052+00:00
RUN ID
a5042cea-6c6b-408d-afdd-a399a385914b
TARGET
Intesa Sanpaolo
STARTED
2026-05-16T10:00:00+00:00
FINAL PHASE
approved
This document is a structural summary of the run's audit trail. It is not a legal opinion and does not certify compliance with any specific regulation. It is intended to support, not replace, review by a qualified compliance professional.

Art. 12Record-keeping

EU AI Act Article 12 requires that high-risk AI systems maintain automatic logs of operations enabling traceability and post-market monitoring. This section reports on the run's logging completeness and tamper-evidence.
State snapshots
1
point-in-time state captures
Approval records
2
human decisions recorded
Notification attempts
1
Slack / email / null fan-out
Hash chain
Chain verified
All 4 entries are linked and self-consistent. This detects accidental corruption or edits made without the chaining code, but NOT an adversary with write access who re-runs the chaining. External audit needs signatures/HMAC, WORM/append-only storage, and/or external anchoring.

Art. 13Transparency & provenance

EU AI Act Article 13 requires that high-risk AI systems be designed and developed in a way that ensures their operation is sufficiently transparent. For an outreach pipeline, this translates into per-output provenance: every generated assertion must be traceable to a verifiable source.
Citation coverage
100.0%
3/3 claims reference at least one citation
Dossier coverage
100.0%
2/2 dossiers carry backing claim IDs
Draft coverage
100.0%
2/2 drafts carry backing claim IDs
Source citations
3
primary sources consulted

Art. 14Human oversight

EU AI Act Article 14 requires that high-risk AI systems be designed and developed in such a way that they can be effectively overseen by natural persons. For outreach automation, this means that no communication leaves the system without an identifiable human decision.
HITL coverage
100.0%
2/2 drafts decided by an identified reviewer
Reviewers
1
distinct identities recorded
Decisions logged
2
across all artifact references
Time to approval
1.0 s
first request to last decision

Decision distribution

Identified reviewers

Generated from a5042cea-6c6b-408d-afdd-a399a385914b.jsonl. All metrics derive from the run's audit trail. The underlying JSONL is the canonical record; this HTML is a rendered view.