Intended Purpose Declaration

Defines the exact environment and conditions under which the AI system is designed to function. This is not a general capability statement but a rigid boundary.

• User Profile: Specifies the intended user (e.g., 'trained radiologist,' not 'general public').
• Deployment Environment: Defines the technical and physical setting (e.g., 'hospital PACS network,' 'cloud API endpoint').
• Interaction Modality: Describes the interface type (e.g., 'API call,' 'graphical user interface').
• Exclusion of Foreseeable Misuse: Explicitly lists prohibited uses that are technically possible but outside the design scope.

A precise description of what the system does, not how it does it. This defines the medical, safety, or legal purpose that triggers specific regulatory annexes.

• Core Task: The primary function (e.g., 'detecting bone fractures,' 'scoring credit risk').
• Output Type: The nature of the result (e.g., 'binary classification,' 'probability score,' 'generated text summary').
• Degree of Autonomy: Specifies if the output is a decision or a suggestion for human override.
• Performance Metrics: The claimed accuracy or error rate that forms the basis of the Conformity Assessment.

The declaration directly determines if the system falls under High-Risk AI System classification per the EU AI Act Annex III. Ambiguity here is a compliance liability.

• Annex III Mapping: The intended purpose must be checked against the list of high-risk use cases (e.g., biometrics, critical infrastructure, education, employment).
• Safety Component Status: Declares if the AI is a safety component of a regulated product (e.g., medical device, machinery).
• General-Purpose AI Distinction: Clarifies if the system is a narrow tool or a General-Purpose AI Registration model with systemic risk potential.

The declaration serves as the baseline against which future changes are measured. Any deviation triggers a Substantial Modification review.

• Change Control Baseline: The original intended purpose is frozen in the Technical Documentation File.
• Re-evaluation Triggers: Expanding the user base, changing the input data type, or altering the output action constitutes a new intended purpose.
• Re-registration Obligation: A substantial modification voids the existing Declaration of Conformity and requires a new Unique Registration ID.

The declaration links the legal entity to the technical artifact, creating a chain of accountability for market surveillance authorities.

• Unique Registration ID: The declaration is permanently bound to the system's ID in the EU AI Act Database.
• Incident Reporting Linkage: Post-market incidents are traced back to the specific intended purpose to determine if the malfunction occurred within the defined scope.
• Supply Chain Responsibility: The Importer Compliance Gate relies on the declaration to verify that the foreign manufacturer's claims match the Union's regulatory requirements.

The declaration forces the provider to articulate the limits of safety, forming the basis for the mandatory Residual Risk Disclosure.

• Known Limitations: The declaration must transparently state what the system cannot do reliably.
• User Information: The intended purpose must be communicated clearly to the deployer to ensure Human Oversight Mechanisms are calibrated correctly.
• Post-Market Monitoring Plan: The declared purpose defines the metrics that must be tracked continuously to detect performance drift outside the specified envelope.