Key Takeaways
| Point |
Summary |
| Shared foundations |
Risk management, documentation, and governance overlap significantly |
| Not sufficient alone |
Existing certifications help but do not guarantee AI Act compliance |
| AI-specific gaps |
Conformity assessment, AI transparency, and technical safety testing are new |
| Efficient integration |
Extending existing frameworks is more efficient than building parallel systems |
| Complementary value |
The combination provides stronger overall governance |
Quick Answer: ISO 27001 and SOC 2 provide valuable foundations for EU AI Act compliance, particularly around risk management, documentation, and governance. However, they do not cover AI-specific requirements like conformity assessment, AI transparency obligations, or technical safety testing. Organizations should extend existing frameworks rather than create separate AI compliance programs.
How Existing Frameworks Help
ISO 27001 Contribution
ISO 27001 establishes an Information Security Management System (ISMS) with practices that support AI Act compliance:
| ISO 27001 Element |
AI Act Benefit |
| Risk assessment process |
Framework for AI risk management system |
| Risk treatment methodology |
Approach for AI risk mitigation |
| Documentation requirements |
Culture and templates for AI documentation |
| Internal audit program |
Process applicable to AI compliance verification |
| Management review |
Governance structure for AI oversight |
| Continual improvement |
Mechanism for AI compliance enhancement |
| Asset inventory |
Foundation for AI system inventory |
| Supplier relationships |
Controls for AI vendor management |
| Incident management |
Process adaptable for AI incidents |
| Competence requirements |
Framework for AI literacy requirements |
SOC 2 Contribution
SOC 2 Trust Services Criteria provide controls relevant to AI Act compliance:
| SOC 2 Criteria |
AI Act Benefit |
| CC1: Control Environment |
Foundation for AI governance |
| CC2: Communication and Information |
Supports transparency requirements |
| CC3: Risk Assessment |
Approach for AI risk identification |
| CC4: Monitoring Activities |
Basis for AI post-market monitoring |
| CC5: Control Activities |
Framework for AI controls |
| CC6: Access Controls |
Security for AI systems |
| CC7: System Operations |
Operational controls for AI |
| CC8: Change Management |
Process for AI system changes |
| CC9: Risk Mitigation |
Approach for AI risk response |
| Processing Integrity |
Supports accuracy requirements |
| Availability |
Supports robustness requirements |
Gap Analysis: What Existing Frameworks Do Not Cover
While ISO 27001 and SOC 2 provide foundations, AI Act compliance requires additional elements:
AI-Specific Technical Requirements
| AI Act Requirement |
Gap Explanation |
| AI risk classification |
Frameworks do not classify systems by AI-specific risk categories |
| Conformity assessment |
No equivalent pre-market assessment procedure |
| CE marking |
Not applicable to information security frameworks |
| Technical documentation for AI |
AI-specific documentation requirements go beyond general IT documentation |
| Data governance for training |
Focus on training data quality and representativeness is AI-specific |
| Accuracy testing |
AI model accuracy testing is not covered by security frameworks |
| Bias assessment |
Frameworks do not require bias or discrimination testing |
AI-Specific Transparency
| AI Act Requirement |
Gap Explanation |
| AI disclosure |
No requirement to disclose AI system use to individuals |
| Synthetic content labeling |
Frameworks do not address AI-generated content |
| Emotion recognition disclosure |
Not covered by security frameworks |
| High-risk AI user information |
Specific information requirements for high-risk AI users |
AI-Specific Governance
| AI Act Requirement |
Gap Explanation |
| Human oversight design |
Technical requirements for human control over AI |
| EU database registration |
No equivalent registration requirement |
| Post-market monitoring |
AI-specific ongoing monitoring obligations |
| Serious incident reporting |
Different from general security incident reporting |
Integration Approach
Rather than building separate compliance programs, extend existing frameworks:
Extended Risk Management
| Current Practice |
AI Act Extension |
| Information security risk assessment |
Add AI-specific risk factors and impact criteria |
| Risk register |
Include AI systems with AI-specific risk attributes |
| Risk treatment plans |
Add AI-specific mitigation measures |
| Risk acceptance |
Define AI risk acceptance criteria and thresholds |
Practical implementation:
- Add AI risk classification as a risk attribute
- Include AI-specific threat scenarios (bias, accuracy degradation, adversarial attacks)
- Define AI impact categories (fundamental rights, safety, discrimination)
- Establish AI risk thresholds aligned with AI Act classifications
Extended Documentation
| Current Practice |
AI Act Extension |
| Policy documentation |
Add AI-specific policies (AI ethics, transparency) |
| Procedure documentation |
Add AI development and deployment procedures |
| Technical documentation |
Add AI-specific elements (training data, model architecture) |
| Records retention |
Include AI-specific records (logs, assessments) |
Practical implementation:
- Create AI policy annex or standalone AI policy
- Add AI system documentation template aligned with Annex IV requirements
- Establish AI documentation lifecycle management
- Define retention periods for AI-specific records
Extended Governance
| Current Practice |
AI Act Extension |
| Management review |
Include AI compliance status |
| Internal audit |
Add AI compliance audit scope |
| Roles and responsibilities |
Define AI-specific responsibilities |
| Competence management |
Add AI literacy requirements |
Practical implementation:
- Add AI to management review agenda with specific AI metrics
- Develop AI compliance audit checklist
- Define AI governance roles (AI compliance lead, system owners)
- Implement AI literacy training program
Extended Vendor Management
| Current Practice |
AI Act Extension |
| Vendor assessment |
Add AI-specific assessment criteria |
| Contractual requirements |
Add AI Act contractual clauses |
| Ongoing monitoring |
Monitor AI vendor compliance |
| Incident coordination |
Establish AI incident coordination with vendors |
Practical implementation:
- Add AI provider assessment questionnaire
- Include AI Act compliance warranties in contracts
- Request conformity assessment evidence from high-risk AI providers
- Define AI incident notification and response protocols
Mapping Frameworks to AI Act
ISO 27001 to AI Act Mapping
| ISO 27001 Control |
AI Act Requirement |
Gap |
| A.5.1 Policies |
General policy framework |
Add AI-specific policies |
| A.5.9 Asset inventory |
AI system inventory |
Add AI classification attributes |
| A.5.23 Supplier relationships |
AI vendor management |
Add AI-specific requirements |
| A.5.24 Incident management |
AI incident handling |
Add AI-specific criteria |
| A.6.3 Awareness and training |
AI literacy |
Add AI-specific training |
| A.8.1 User endpoint devices |
Human oversight interfaces |
Add AI-specific interface requirements |
| A.8.9 Configuration management |
AI system configuration |
Add AI-specific documentation |
| A.8.16 Monitoring activities |
Post-market monitoring |
Add AI-specific monitoring |
| A.8.32 Change management |
AI system changes |
Add re-classification assessment |
SOC 2 to AI Act Mapping
| SOC 2 Control Point |
AI Act Requirement |
Gap |
| CC1.1 Entity commitment |
AI governance commitment |
Add explicit AI governance |
| CC2.2 Internal communication |
AI transparency |
Add AI disclosure requirements |
| CC3.1 Risk identification |
AI risk classification |
Add AI-specific risk criteria |
| CC3.4 Risk assessment |
AI risk management system |
Add AI-specific methodology |
| CC4.2 Ongoing evaluations |
Post-market monitoring |
Add AI-specific monitoring |
| CC5.2 Control selection |
AI controls |
Add AI-specific controls |
| CC7.2 System monitoring |
AI operation monitoring |
Add AI behavior monitoring |
| CC8.1 Change authorization |
AI change assessment |
Add re-classification triggers |
| PI1.1 Data quality |
Training data governance |
Add AI-specific data requirements |
Benefits of Integration
Extending existing frameworks rather than building separate programs offers several advantages:
| Benefit |
Explanation |
| Efficiency |
Leverage existing processes, documentation, and governance |
| Consistency |
Single integrated approach rather than competing systems |
| Resource optimization |
Use existing teams and tools |
| Reduced complexity |
One framework to maintain rather than multiple |
| Audit efficiency |
Integrated audits covering multiple requirements |
| Stakeholder clarity |
Clear ownership and accountability |
Implementation Considerations
For Organizations with ISO 27001
| Step |
Action |
| 1 |
Add AI to scope of ISMS |
| 2 |
Extend risk assessment methodology for AI |
| 3 |
Add AI controls to Statement of Applicability |
| 4 |
Update documentation to include AI elements |
| 5 |
Add AI to internal audit program |
| 6 |
Include AI in management review |
| 7 |
Consider ISO 42001 for comprehensive AI management |
For Organizations with SOC 2
| Step |
Action |
| 1 |
Add AI systems to scope |
| 2 |
Extend control descriptions to cover AI |
| 3 |
Add AI-specific policies and procedures |
| 4 |
Include AI in testing procedures |
| 5 |
Update system description for AI elements |
| 6 |
Coordinate with auditor on AI scope |
For Organizations with Both
| Step |
Action |
| 1 |
Identify overlapping AI requirements across frameworks |
| 2 |
Create unified AI governance approach |
| 3 |
Develop integrated AI documentation |
| 4 |
Establish single AI risk assessment process |
| 5 |
Coordinate audit timing and scope |
ISO 42001: The AI-Specific Standard
ISO 42001 provides a comprehensive AI management system framework that complements ISO 27001 and aligns well with EU AI Act requirements:
| ISO 42001 Element |
AI Act Alignment |
| AI policy and objectives |
Governance requirements |
| AI risk assessment |
Risk management system |
| AI impact assessment |
Fundamental rights assessment |
| AI system lifecycle management |
Documentation requirements |
| Data management for AI |
Data governance requirements |
| AI transparency |
Transparency obligations |
| Third-party relationships |
Vendor management |
| Monitoring and measurement |
Post-market monitoring |
Organizations serious about AI governance may consider ISO 42001 certification as a comprehensive approach that supports AI Act compliance.
How Bastion Helps
Bastion helps organizations leverage existing compliance programs for AI Act compliance:
- Framework assessment. We evaluate your current ISO 27001 or SOC 2 implementation against AI Act requirements.
- Gap identification. We identify specific AI-related gaps in your existing frameworks.
- Integration planning. We design an efficient approach to extend existing frameworks for AI compliance.
- Documentation support. We help update policies, procedures, and documentation for AI requirements.
- Audit coordination. We help prepare for integrated audits covering multiple frameworks.
- ISO 42001 readiness. We assess readiness for ISO 42001 certification if desired.
Our experience with ISO 27001 and SOC 2 compliance informs our approach to AI Act integration.
Ready to extend your existing compliance programs for the AI Act? Talk to our team
Sources