Live demo of the Axiom Audit Portal. Data is synthetic but representative of a real assessment.
Demo AI Platform — EU AI Act Preview
Recommended Assessment Outcome
BLOCK
Residual Non-Conformity Index
MODERATE
Score: 0.55
(Unwitnessed ÷ Total) × Severity × Staleness
Passing
5
of 7 requirements
Findings
1
open findings
Risk Classification
Annex IIIEU AI Act risk tier determines which conformity obligations apply. Select the classification that matches this AI system.
AI systems in Annex III areas that pose significant risks to health, safety, or fundamental rights. Requires third-party or internal conformity assessment.
Obligations
Full conformity assessment required: risk management (Art. 9), data governance (Art. 10), logging (Art. 12), transparency (Art. 13), human oversight (Art. 14), accuracy & robustness (Art. 15), post-market monitoring (Art. 72).
Annex III Categories
EU AI Act Conformity
Regulation (EU) 2024/1689Loading risk register...
Loading checklist...
Regulation Articles
AI Governance
Requirements Details
| Procedure | Family | Title | Severity | Verdict | Last Witnessed | Verify |
|---|---|---|---|---|---|---|
| AI-GRD.2 | AI | Content Safety Filter — Output Classification Passed Clean Expected: 0.5Refusal Detected: 1 Open 52d | — | FAIL | 4/12/2026 | Verify → |
| AI-CHR.1 | AI | Agent Charter Policy Manifest | — | UNKNOWN | — | — |
| AI-INF.1 | AI | Inference Provenance — Prompt/Response Hash Capture Hash Required: 1Hashes Present: 1 | — | PASS | 4/20/2026 | Verify → |
| AI-INF.2 | AI | Inference Latency — Response Time Within Threshold Latency Limit (ms): 1Actual Latency (ms): 1 | — | PASS | 4/20/2026 | Verify → |
| AI-MDL.2 | AI | Model Version Tracking — Version Identifier Recorded Version Required: 1Version Recorded: 1Context: 1 | — | PASS | 4/8/2026 | Verify → |
| AI-GRD.1 | AI | Guardrail Enforcement — Required Safety Filters Active Required Guardrails: 1Active Guardrails: 1 | — | PASS | 4/20/2026 | Verify → |
| AI-MDL.1 | AI | Model Weight Integrity — Deployed Hash Matches Approved Hash Required: 1Hash Present: 1 | — | PASS | 4/20/2026 | Verify → |
Finding Register
Loading findings...
Evidence Requests
Loading requests...
Assessment Interviews
Loading interviews...
Statistical Sampling
Fisher-YatesSame seed + same population = same sample. Reproducible for peer review.
Plan of Action & Milestones
Open
7
Overdue
0
Closed
55
| Procedure | Severity | Status | Days Open | Milestone |
|---|---|---|---|---|
| AI-GOV.3 | — | OPEN | 50d | 6/13/2026 |
| AI-EXPL.2 | — | OPEN | 51d | 6/12/2026 |
| AI-FAIR.2 | — | OPEN | 51d | 6/12/2026 |
| AI-GRD.2 | — | OPEN | 52d | 6/11/2026 |
| AI-EXPL.1 | — | OPEN | 54d | 6/9/2026 |
| AI-4.2 | — | OPEN | 54d | 6/9/2026 |
| AI-HITL.3 | — | OPEN | 56d | 6/7/2026 |
Framework Coverage
AI Witness Summary
Models Witnessed
7
Total Inferences
146
Pass Rate
58%
Agent Transparency
Art. 13 & 14Complete inventory of AI agents witnessed in this system. EU AI Act Articles 13 (Transparency) and 14 (Human Oversight) require disclosure of agent identity, capabilities, and operational boundaries.
Agents
13
Models
7
Tools Invoked
7
Violations
3
| Agent | Models | Inferences | Pass Rate | Clearing | Violations | Active Period |
|---|---|---|---|---|---|---|
▶ — | claude-haiku-4-5 | 36 | 53% | L0 Analytics | 0 | Mar 31 — Apr 20 |
▶ — | gpt-4o-mini | 28 | 57% | L0 Analytics | 0 | Mar 31 — Apr 20 |
▶ — | gpt-4o | 25 | 64% | L0 Analytics | 0 | Apr 2 — Apr 20 |
▶ — | claude-sonnet-4-6 | 24 | 58% | L0 Analytics | 0 | Mar 31 — Apr 20 |
▶ — | demo-compliance-model-v1 | 12 | 83% | L0 Analytics | 0 | Apr 8 — Apr 15 |
▶ support-bot | claude-sonnet-4-6 | 4 | 25% | L1 Standard | 1 | Apr 18 — Apr 19 |
▶ compliance-agent | a2a69af70d1b9be7 | 3 | 100% | L3 Classified | 0 | Apr 18 — Apr 19 |
▶ research-coordinator | claude-opus-4-6 | 3 | 0% | L0 Analytics | 0 | Apr 18 — Apr 19 |
▶ doc-processor | gpt-4o-mini | 3 | 100% | L2 Sensitive | 0 | Apr 18 — Apr 19 |
▶ — | claude-opus-4-6 | 2 | 0% | L0 Analytics | 0 | Apr 19 — Apr 19 |
▶ risk-assessor | gpt-4o | 2 | 0% | L2 Sensitive | 1 | Apr 19 — Apr 19 |
▶ loan-orchestrator | gpt-4o | 2 | 0% | L2 Sensitive | 1 | Apr 19 — Apr 19 |
▶ fraud-orchestrator | gpt-4o | 2 | 100% | L1 Standard | 0 | Apr 19 — Apr 19 |
Loading monitoring data...
Agent Subway Map
Art. 12 · Art. 14Multi-agent decision lineage with cryptographic proof at every handoff. Hover nodes for human-readable context. Click to drill down.
Chains
4
Total Nodes
8
Agents
4
Violations
1
Ready to assess your AI system?
Get your own audit portal with cryptographic evidence, Annex IV checklist, and assessment tools. Free tier includes 90-day anchor retention and up to 1,000 inferences.
Every finding in this report is backed by a cryptographic SWT3 Witness Anchor. Anchors can be independently verified at any time using the public verification endpoint.
Axiom is a neutral evidence platform. It does not grant certifications, issue conformity assessments, or replace the professional judgment of a qualified auditor, Notified Body, or Authorizing Official. Scores, gate decisions, and recommended outcomes are computed deterministically from machine-gathered evidence and are provided for informational purposes only. The final authority on conformity or compliance rests with the designated assessment body and their professional standards.
SWT3 Protocol, Witness Anchor format, UCT Registry, and AI Witness clearing methodology are Patent Pending (USPTO Provisional Patent Applications filed 2026). © 2026 TeNova Labs. All rights reserved.