Harmonised standards are EU-recognised technical specifications developed by standardisation bodies such as CEN, CENELEC, and ETSI. Under the EU AI Act, these standards provide a presumption of conformity with specific legal obligations, especially for high-risk AI systems. By aligning with harmonised standards, AI providers can streamline compliance, reduce regulatory uncertainty, and demonstrate adherence to the Act’s core requirements for safety, robustness, transparency, and risk management.
1. Background and Establishment
The EU has long relied on harmonised standards to implement complex legislation in a technically neutral, industry-driven way. For AI, this approach continues under the EU Artificial Intelligence Act, where harmonised standards provide the practical bridge between high-level legal obligations and operational compliance.
These standards are developed by recognised European standardisation organisations (ESOs), including:
- CEN – European Committee for Standardization
- CENELEC – European Committee for Electrotechnical Standardization
- ETSI – European Telecommunications Standards Institute
Once cited in the Official Journal of the EU (OJEU), these standards become official tools to demonstrate conformity with the AI Act’s requirements.
2. Purpose and Role in the EU AI Ecosystem
Harmonised standards serve several purposes:
- Provide concrete technical benchmarks for AI developers and providers
- Support the conformity assessment process (e.g. CE marking)
- Facilitate cross-border compliance consistency across the EU
- Allow innovation while maintaining legal alignment
- Reduce the regulatory burden on small and medium enterprises by offering clear, structured guidance
They are essential to operationalising the AI Act’s abstract principles like fairness, explainability, and risk mitigation.
3. Legal Foundations in the EU AI Act
The relevance of harmonised standards is grounded in:
- Article 40 – AI systems that comply with harmonised standards are presumed to conform with the relevant requirements of the AI Act
- Article 43–47 – Conformity assessments can rely on these standards to demonstrate compliance, especially under internal control procedures
- Annex IV – Technical documentation must reference applicable harmonised standards used in system design and evaluation
This presumption of conformity does not exempt providers from liability but offers a structured path toward compliance.
4. Connection to the EU AI Safety Alliance
The EU AI Safety Alliance supports organisations in:
- Identifying applicable harmonised standards based on system type and risk level
- Integrating standards like ISO/IEC 42001 (AI management systems) or ISO/IEC 24029 (robustness & accuracy)
- Providing templates, mappings, and compliance roadmaps
- Supporting conformity assessments through gap analysis and certification readiness
By following harmonised standards, organisations can engage with regulators and Notified Bodies more effectively and confidently.
5. Current and Emerging Standards Relevant to the AI Act
Key harmonised and pre-harmonised standards include:
- ISO/IEC 42001 – AI Management System Standard
- CEN-CLC/JTC 21 – Governance and risk management of AI systems
- ISO/IEC 23894 – Risk management for AI
- ISO/IEC TR 24028 – Trustworthiness in AI
- ISO/IEC 24029 – Robustness of neural networks
- EN 301 549 – Accessibility requirements for ICT products and services
As AI-specific standards continue to evolve, the Commission will expand the list of harmonised standards through standardisation requests to the ESOs.
6. Practical Benefits of Using Harmonised Standards
Using harmonised standards provides:
- Clarity – Translates abstract legal concepts into concrete technical requirements
- Efficiency – Reduces the cost and complexity of compliance
- Consistency – Enables uniform practices across development teams and jurisdictions
- Defensibility – Offers a documented rationale for system design choices
- Auditability – Simplifies internal and external conformity assessments
While not legally mandatory, they offer a de facto roadmap for responsible AI deployment.
7. How to Incorporate Harmonised Standards into Your Compliance Program
To leverage harmonised standards:
- Determine whether your AI system is classified as high-risk
- Identify applicable harmonised standards from the EU’s published list
- Map each legal requirement (e.g. risk management, human oversight, robustness) to its corresponding technical standard
- Integrate these requirements into your design, documentation, and testing protocols
- Include references to the standards in your technical documentation (Annex IV)
- Engage with the EU AI Safety Alliance to review your conformity strategy
- Keep up with updates, as standards will evolve alongside technological advances and regulatory shifts
By embedding harmonised standards, you ensure your compliance program is future-proof and regulator-ready.