AI Model Registry
8 models · 3 requiring action
Add model
8
Total models
5
Compliant
2
Needs review
1
Overdue
ModelRisk levelStatusLast doc
High Risk
Needs Review
Nov 3, 2025
High Risk
Compliant
Mar 15, 2026
High Risk
Overdue
Never
Limited
Compliant
Feb 28, 2026
High Risk
Generating…
Jan 20, 2026
ModelsFraudDetect-RT
FraudDetect-RT
High Risk · Annex III §5b
Needs Review
Last documented Nov 3, 2025
Generate documentation
Model info
Versionv2.4.1
FrameworkPyTorch 2.1
OwnerRisk Engineering
DeployedSep 12, 2024
Compliance status
ArticlesArt. 9, 10, 11, 13, 14
Last auditNov 3, 2025
Next dueJun 1, 2026
StatusReview needed
3 documents out of date
Technical Documentation Pack · Risk Assessment Report · Human Oversight Protocol
Generating will run all 4 compliance agents and produce 6 audit-ready documents. Estimated time: 8–12 minutes.
FraudDetect-RTGenerating
Generating documentation
All 4 agents running
0:00
elapsed
Overall progress0%
Compliance agents
Model agentQueued
Waiting to start…
Infrastructure agentQueued
Waiting for model agent…
Governance agentQueued
Waiting for model agent…
Regulatory rules engineQueued
Waiting for evidence aggregation…
FraudDetect-RTDocuments
Generated documents
6 documents · May 18, 2026 · Generated in 0:08
Export all
Needs review
Needs review
Ready
Ready
Ready
Ready
DocumentsTechnical Documentation Pack
Technical Documentation Pack
FraudDetect-RT v2.4.1 · EU AI Act Article 11 · Generated May 18, 2026
1. General description
FraudDetect-RT is a real-time machine learning system deployed within ABN AMRO's transaction processing infrastructure for the purpose of identifying potentially fraudulent financial transactions. The system operates as a High-Risk AI System under Annex III §5(b) of the EU AI Act.
2. Intended purpose
The system is intended to support human analysts in the detection of fraudulent payment transactions across retail and commercial banking products. Output scores are advisory and subject to human review before action is taken.
Flag · Article 14
Human oversight claim requires verification. No linked SOP document found in Confluence.
3. Training data
The model was trained on 24 months of historical transaction data (Jan 2022 – Dec 2023), comprising 847M transactions. Data quality assessments were conducted per Article 10 requirements. No personal data categories beyond transaction metadata were used.
Note · Article 9(7)
Training data cutoff is 17 months old. Consider whether re-training is required under continuous monitoring obligations.
Flags 2
Human oversight SOP missing
Art. 14 requires a documented oversight procedure linked to this model.
Training data age
Data is 17 months old — review Art. 9(7) continuous monitoring.
Approve
Request changes
Export PDF
Documentation approved
Technical Documentation Pack for FraudDetect-RT has been approved and logged to the audit trail.
Audit trail entry
Approved byS. Agarwal
TimestampMay 18, 2026 · 14:32
ModelFraudDetect-RT v2.4.1
Documents6 generated · 1 flagged
Export all as PDF
Back to dashboard