The self-assessment path used for most Annex III high-risk AI systems. The provider verifies its own quality management system, prepares technical documentation, and registers the system directly, without an external notified body unless the rules require one.
What the Internal Control Procedure Actually Does
The internal control procedure is the self-assessment route for most Annex III high-risk AI systems. The provider checks its own quality management system, prepares the technical file, and registers the system directly unless a notified body is required.
This matters because it is not just a filing step. It is the compliance path that places primary responsibility on the provider to prove conformity, maintain the underlying controls, and produce the evidence expected by the EU AI Act.
Where It Sits in the EU AI Act Conformity Process
For most high-risk systems, the internal control procedure is the default conformity assessment route. It sits between design and market entry, turning the provider’s internal governance, testing, and documentation into the basis for declaration of conformity.
That structure makes the procedure especially important for Annex III systems, where the legal question is often not whether controls exist in theory, but whether they have been implemented, verified, and documented well enough to support self-assessment.
The difference from an external assessment is practical as well as legal. A provider can move faster with internal control, but it also carries the burden of making sure its evidence is complete enough to withstand regulatory scrutiny later.
What Providers Must Be Able to Show
The procedure assumes that the provider can evaluate its own controls in a disciplined way. That means the quality management system, technical documentation, risk management evidence, and post-market processes all need to align, not just exist as separate artifacts.
A useful way to think about the procedure is as an internal proof package. If the system is challenged by a regulator, customer, or auditor, the provider needs to show how the system was designed, tested, controlled, and approved under a documented process rather than ad hoc judgment.
That is why it also connects to broader control disciplines such as segregation of duties and internal governance. NHIMG’s Segregation of Duties (SoD) Guide is a useful companion where the same organisation needs to prevent one team from designing, approving, and releasing controls without challenge.
How It Affects Evidence, Accountability, and Market Entry
The internal control procedure compresses accountability into the provider’s own organisation, which is efficient but unforgiving. If the technical documentation is thin, the QMS is weak, or the conformity basis is unclear, the problem does not disappear because no external body was involved.
In practice, the procedure also shapes how teams work across legal, compliance, engineering, and product functions. The provider must be able to register confidently, support the declaration of conformity, and keep the evidence current as the system changes.
For readers mapping the legal obligation back to primary source text, the EU AI Act regulatory framework is the canonical reference for the high-risk system regime and its conformity pathways.
Risk and Threat Considerations
The main risk is that self-assessment can create a false sense of completeness. If the provider’s internal controls are weak, the procedure can turn into paper compliance, where documentation looks formal but does not accurately reflect how the system was built, tested, or governed.
Failure mechanism: Gaps in quality management, undocumented design decisions, weak evidence trails, or poorly controlled changes can lead to a flawed conformity assessment that survives internally but fails under external review.
Impact: The result can be delayed market access, regulatory findings, forced remediation, and loss of trust in the system’s declared compliance posture.
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
NIST CSF 2.0 and NIST SP 800-53 Rev 5 set the technical controls, while EU AI Act, ISO/IEC 27001:2022 and SOC 2 (AICPA) define the regulatory obligations.
| Framework | Control / Reference | Relevance |
|---|---|---|
| EU AI Act | High-risk AI system conformity assessment and internal control procedure | Governs the self-assessment route for Annex III high-risk AI systems. |
| Recommendation — Use the internal control procedure to verify QMS, technical documentation, and conformity before registration. | ||
| NIST CSF 2.0 | GV.PO-01 — Policy | The procedure depends on documented organisational policy and accountability for compliance evidence. |
| Recommendation — Define policy ownership for AI conformity evidence and registration decisions. | ||
| ISO/IEC 27001:2022 | A.5.1 — Policies for information security | The procedure relies on documented internal policies and governed evidence processes. |
| Recommendation — Maintain documented policies that support repeatable internal compliance assessment. | ||
| NIST SP 800-53 Rev 5 | PM-10 — System Development Life Cycle | The procedure depends on disciplined lifecycle evidence for controlled system release. |
| Recommendation — Embed conformity evidence into the system lifecycle before registration. | ||
| SOC 2 (AICPA) | CC2.1 — Commitment to Integrity and Ethical Values | Self-assessment relies on accountable control ownership and reliable internal reporting. |
| Recommendation — Assign clear accountability for control evidence and formal approval. | ||
Practitioner Guidance
Governance implication: Treat the internal control procedure as a control discipline, not a document collection exercise. Ownership should be explicit across product, compliance, and engineering so that the QMS, technical documentation, and registration evidence remain consistent as the system evolves.
What to watch for: The strongest warning sign is a mismatch between what the organisation can actually demonstrate and what the conformity file says. If the evidence depends on informal assumptions or unowned approvals, the procedure is not ready for dependable self-assessment.
Related resources from NHI Mgmt Group
Deepen Your Knowledge
Free weekly newsletter
Subscribe to the NHI & AI Identity Journal
The latest on NHI and Agentic AI security – articles, research, breaches, news and events every week.
Bonus 33% off our NHI Course when you subscribe.
Reviewed and updated by the NHIMG editorial team on October 12, 2026.
NHI Mgmt Group — the #1 independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org