Quality Management System (QMS) Templates
EU AI Act Article 17 requires providers of high-risk AI systems to establish a Quality Management System. These templates provide a starting point — adapt them to your organization.
Template
EU AI Act Reference
Purpose
Model Training SOP
Art. 17(1)(b)(c)(d)
Standard procedure for fine-tuning approval, execution, and validation
Data Management SOP
Art. 17(1)(f), Art. 10
Data collection, annotation, quality assurance, and governance
Incident Response SOP
Art. 17(1)(h)(i)
Handling model failures, safety incidents, and corrective actions
Change Management SOP
Art. 17(1)(b)(c)
Versioning, review, approval, and rollback procedures
Roles & Responsibilities
Art. 17(1)(m)
AI Officer, Data Steward, ML Engineer, Compliance Officer roles
Access Control
Art. 17(1)(c) + ISO A.5.15-A.8.5
Operator identity + secret rotation (Wave 4 / Faz 23)
Encryption at Rest
Art. 17(1)(c) + ISO A.5.33 + A.8.10 + A.8.24
Substrate-side encryption guidance (Wave 4 / Faz 23)
Risk Treatment Plan
Art. 17(1)(c) + ISO A.5.7-A.8.30
ISO 27005 risk register template (Wave 4 / Faz 23)
Statement of Applicability
Art. 17(1)(c) + ISO 6.1.3 d)
93-control applicability matrix (Wave 4 / Faz 23)
Copy these templates to your internal documentation system
Replace [YOUR ORGANIZATION] with your company name
Assign real names to each role
Review and approve with your legal/compliance team
Reference ForgeLM's automated artifacts as evidence
ForgeLM Automated Artifacts (Maps to QMS)
QMS Requirement
ForgeLM Artifact
Generated By
Training records
compliance_report.json
forgelm --config job.yaml
Data provenance
data_provenance.json
Automatic per run
Evaluation evidence
benchmark_results.json, safety_results.json
Automatic per run
Model identity
model_integrity.json
Automatic per run
Audit trail
audit_log.jsonl
Automatic per run
Risk assessment
risk_assessment.json
From config risk_assessment: section
Deployer instructions
deployer_instructions.md
Automatic per run