Inferensys

Glossary

Declaration of Conformity

The legally binding document signed by the provider asserting that a high-risk AI system satisfies all applicable regulatory requirements, required for database registration.
Data scientist building training data pipeline on laptop, data preprocessing visible, technical workspace.
REGULATORY COMPLIANCE

What is a Declaration of Conformity?

The legally binding document asserting a high-risk AI system meets all applicable EU AI Act requirements.

A Declaration of Conformity is a formal legal instrument issued and signed by the provider of a high-risk AI system, explicitly asserting that the system satisfies all essential requirements of the EU AI Act. By drafting this document, the provider assumes full legal responsibility for the system's compliance with applicable harmonized standards and regulatory obligations, enabling the affixing of the CE marking.

This declaration must contain specific information, including the system's unique registration ID, the provider's name and address, and a reference to the relevant conformity assessment procedure performed. It serves as the primary evidentiary link between the physical or digital product and its entry in the EU AI Act Database, and must be kept up-to-date following any substantial modification.

ANATOMY OF A LEGALLY BINDING ASSERTION

Essential Components of the Declaration

The Declaration of Conformity is a formal legal instrument, not merely a technical summary. It must contain specific mandatory elements to be valid under the EU AI Act, serving as the provider's sworn attestation of compliance.

01

Provider Identification & Mandate

The declaration must unambiguously identify the provider (manufacturer) and, if applicable, their authorized representative established within the Union. This includes the registered trade name, address, and the explicit statement that the declaration is issued under the sole responsibility of the provider. For non-EU entities, the Authorized Representative Mandate must be cited, establishing the legal nexus for enforcement within the jurisdiction.

Art. 47
EU AI Act Provision
02

System Identification & Traceability

A precise, unambiguous identification of the AI system is required to link the physical or digital product to its compliance record. This includes:

  • The model name and trade name of the system.
  • The Unique Registration ID assigned by the EU database upon successful registration.
  • Any additional serial numbers or software versioning that allows traceability throughout the product's lifecycle. This block creates an unbreakable chain between the physical CE marking and the digital Technical Documentation File.
03

Conformity Statement & Standards

This is the core performative clause where the provider explicitly states that the AI system is in conformity with the EU AI Act and any other applicable Union harmonization legislation. It must reference:

  • The specific Conformity Assessment procedure followed (e.g., internal control or Notified Body review).
  • The Harmonized Standards or common specifications applied to achieve the Presumption of Conformity.
  • The identification number of the Notified Body that issued the EU-type examination certificate, if applicable.
04

Intended Purpose & Residual Risk

The declaration legally binds the system to its Intended Purpose Declaration, defining the exact use case and operational context. It must transparently disclose any Residual Risk Disclosure—risks that could not be fully mitigated through design. This section forms the legal boundary of the registration; any use outside this stated purpose, or a Substantial Modification to the system's logic, immediately voids the declaration and triggers a mandatory re-assessment.

05

Signature & Legal Force

The declaration is not valid until signed. It requires the place and date of issue and the signature, name, and function of the person empowered to bind the provider. This transforms the document from a technical report into a legally binding affidavit. A false declaration triggers the full penalty mechanisms of the Act, including potential Registration Suspension and Market Withdrawal Notification by the National Competent Authority.

06

Multi-Lingual Translation Requirement

To ensure accessibility for end-users and market surveillance authorities, the Declaration of Conformity must be translated into the official language(s) required by the Member State where the product is placed on the market. This is not merely a courtesy; it is a legal prerequisite for valid CE Marking. The translated version carries the same legal weight as the original and must be maintained as part of the Post-Market Monitoring documentation package.

DECLARATION OF CONFORMITY

Frequently Asked Questions

A Declaration of Conformity is the legally binding document signed by the provider asserting that a high-risk AI system satisfies all applicable regulatory requirements. It is a prerequisite for CE marking and mandatory database registration under the EU AI Act.

A Declaration of Conformity (DoC) is a legally binding document issued by the provider of a high-risk AI system, explicitly asserting sole responsibility that the system satisfies all applicable essential requirements of the EU AI Act. It serves as the formal conclusion of the conformity assessment procedure. The DoC must contain the system's unique registration ID, the provider's name and address, a statement of sole responsibility, and a reference to the relevant harmonized standards applied. Without a properly executed DoC, a CE marking cannot be affixed, and the system cannot be registered in the EU AI Act Database or placed on the Union market.

Prasad Kumkar

About the author

Prasad Kumkar

CEO & MD, Inference Systems

Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.

His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.