Skip to content

ARAL Privacy v1.1

Agent Reference Architecture Layers — Privacy & GDPR Compliance Specification

Version: 1.1.0
Status: Normative
Date: 2026-01-15
License: CC-BY-4.0


This specification defines privacy requirements and GDPR compliance mechanisms for ARAL-conformant agents. It provides normative requirements for personal data handling, data subject rights implementation, and privacy-by-design principles.

Keywords: ARAL, privacy, GDPR, data protection, consent management, data subject rights


  1. Introduction
  2. Scope
  3. Normative References
  4. Terms and Definitions
  5. Privacy Principles
  6. Personal Data Handling
  7. Data Subject Rights
  8. Consent Management
  9. Cross-Border Data Transfer
  10. Privacy Architecture
  11. Conformance

This specification extends ARAL-CORE-1.0 and ARAL-SECURITY-1.0 to provide comprehensive privacy and GDPR compliance requirements for AI agents handling personal data within the European Union and other jurisdictions with data protection laws.

1.2 Relationship to Other ARAL Specifications

Section titled “1.2 Relationship to Other ARAL Specifications”
ARAL-PRIVACY-1.0
├─ Extends: ARAL-CORE-1.0
├─ Requires: ARAL-SECURITY-1.0
└─ Referenced by: ARAL-CONFORMANCE-1.0
  • Privacy by Design: Embed privacy into architecture from the start
  • Data Minimization: Collect only necessary personal data
  • Transparency: Clear disclosure of data practices
  • User Control: Enable meaningful consent and preferences
  • Accountability: Demonstrate compliance through technical measures

  • Personal data collection, processing, and storage requirements
  • Implementation of all GDPR data subject rights
  • Consent management mechanisms
  • Cross-border data transfer controls
  • Privacy breach detection and notification
  • Data retention and deletion policies
  • Legal interpretation of GDPR (consult legal counsel)
  • Business process compliance (DPO responsibilities, etc.)
  • Specific jurisdictional requirements beyond GDPR
  • Marketing and analytics platform integration

The following documents are referenced in this specification:

  • GDPR — Regulation (EU) 2016/679 General Data Protection Regulation
  • ePrivacy Directive — Directive 2002/58/EC (as amended)
  • ARAL-CORE-1.0 — ARAL Core Specification
  • ARAL-SECURITY-1.0 — ARAL Security Specification
  • ISO/IEC 27701 — Privacy Information Management System
  • ISO/IEC 29134 — Privacy Impact Assessment Guidelines
  • RFC 2119 — Key words for use in RFCs to Indicate Requirement Levels

Information relating to an identified or identifiable natural person (GDPR Art. 4(1)).

Natural person whose personal data is processed (GDPR Art. 4(1)).

Entity determining purposes and means of personal data processing (GDPR Art. 4(7)).

Entity processing personal data on behalf of controller (GDPR Art. 4(8)).

Any operation performed on personal data (GDPR Art. 4(2)).

Freely given, specific, informed, and unambiguous indication of data subject’s wishes (GDPR Art. 4(11)).

Breach of security leading to accidental or unlawful destruction, loss, alteration, or unauthorized disclosure of personal data (GDPR Art. 4(12)).


Agents MUST implement privacy protection measures during design and development, not as an afterthought.

Implementation: Integrate privacy controls at all architecture layers.

Agents MUST apply strictest privacy settings by default, requiring explicit user action to reduce privacy protection.

Implementation: Default to minimal data collection, shortest retention, strictest access controls.

Agents MUST collect only personal data that is adequate, relevant, and necessary for specified purposes (GDPR Art. 5(1)(c)).

Implementation:

  • Document purpose for each data element
  • Reject unnecessary data collection
  • Regular data minimization audits

Agents MUST process personal data only for specified, explicit, and legitimate purposes (GDPR Art. 5(1)(b)).

Implementation: Tag all data with processing purpose; prevent purpose creep.

Agents MUST retain personal data no longer than necessary for processing purposes (GDPR Art. 5(1)(e)).

Implementation: Automated retention policies with deletion triggers.

Agents MUST ensure personal data is accurate and kept up to date (GDPR Art. 5(1)(d)).

Implementation: Data validation, correction workflows, staleness detection.

[REQ-PRIV-007] Integrity and Confidentiality

Section titled “[REQ-PRIV-007] Integrity and Confidentiality”

Agents MUST process personal data securely, protecting against unauthorized access, loss, or damage (GDPR Art. 5(1)(f)).

Implementation: Encryption, access controls, audit logging (see ARAL-SECURITY-1.0).


[REQ-PRIV-008] Sensitive Data Identification

Section titled “[REQ-PRIV-008] Sensitive Data Identification”

Agents MUST identify and classify special categories of personal data (GDPR Art. 9):

  • Racial or ethnic origin
  • Political opinions
  • Religious or philosophical beliefs
  • Trade union membership
  • Genetic data
  • Biometric data
  • Health data
  • Sex life or sexual orientation

Implementation: Metadata tagging system with sensitivity levels.

Agents MUST provide clear, accessible privacy notice before collecting personal data (GDPR Art. 13-14).

Required Information:

  • Controller identity and contact
  • Data Protection Officer (DPO) contact
  • Processing purposes and legal basis
  • Data recipients
  • Retention period
  • Data subject rights
  • Right to withdraw consent
  • Right to lodge complaint with supervisory authority

Agents MUST establish lawful basis for processing before collecting personal data (GDPR Art. 6):

  1. Consent
  2. Contract performance
  3. Legal obligation
  4. Vital interests
  5. Public task
  6. Legitimate interests

Implementation: Document and track legal basis per processing activity.

Personal data MUST be encrypted at rest using approved algorithms (see ARAL-SECURITY-1.0 §5.3).

Personal data MUST be encrypted in transit using TLS 1.3 or higher (see ARAL-SECURITY-1.0 §5.4).

Personal data belonging to different data subjects MUST be logically or physically isolated.

Implementation: Multi-tenant architecture with tenant isolation controls.


ARAL agents MUST implement technical mechanisms to facilitate all GDPR data subject rights.

Agents MUST provide data subjects with access to their personal data within 30 days.

Response Format:

{
"request_id": "uuid",
"data_subject": "user@example.com",
"timestamp": "2026-01-15T14:30:00Z",
"data": {
"profile": { ... },
"interactions": [ ... ],
"preferences": { ... }
},
"processing_purposes": [ "service_delivery", "analytics" ],
"recipients": [ "processor_a", "processor_b" ],
"retention_periods": { "profile": "5 years", "logs": "90 days" },
"rights": [ "access", "rectification", "erasure", ... ]
}

Agents MUST provide mechanisms to correct inaccurate personal data within 30 days.

Implementation: RESTful API endpoint, UI correction forms, versioned updates.

Agents MUST delete personal data when:

  • No longer necessary for original purpose
  • Consent withdrawn (and no other legal basis)
  • Data subject objects (and no overriding legitimate grounds)
  • Data unlawfully processed
  • Legal obligation requires deletion
  • Data collected from children (Art. 8)

Exceptions:

  • Legal obligation to retain
  • Public interest or official authority
  • Legal claims defense
  • Archiving in public interest

Implementation:

deletion_workflow:
- verify_identity
- check_exceptions
- cascade_delete_across_systems
- anonymize_logs
- generate_confirmation
- audit_trail

Agents MUST delete or anonymize personal data across all systems, backups, and third-party processors.

Implementation: Event-driven deletion propagation, backup sanitization.

Agents MUST provide personal data in structured, commonly used, machine-readable format (JSON, XML, CSV).

Export Formats:

  • JSON (preferred)
  • XML
  • CSV
  • Industry-standard formats (e.g., HL7 FHIR for health data)

7.5 Right to Restriction of Processing (Art. 18)

Section titled “7.5 Right to Restriction of Processing (Art. 18)”

Agents MUST support restriction of processing when:

  • Accuracy contested
  • Processing unlawful but data subject opposes deletion
  • Controller no longer needs data but data subject needs it for legal claims
  • Data subject objects pending verification of legitimate grounds

Implementation: Status flag on data records preventing further processing except storage.

Agents MUST allow data subjects to object to processing based on legitimate interests or direct marketing.

Implementation: Opt-out controls, granular processing preferences.

Section titled “7.7 Rights Related to Automated Decision-Making (Art. 22)”

Agents MUST NOT make solely automated decisions with legal or significant effects without:

  • Explicit consent, or
  • Contract necessity, or
  • Legal authorization

AND must provide:

  • Right to human intervention
  • Right to express viewpoint
  • Right to contest decision

Implementation: Human-in-the-loop workflows, explainability features.


Agents MUST obtain consent that is:

  • Freely given: No coercion, genuine choice
  • Specific: Clear purpose stated
  • Informed: Comprehensive information provided
  • Unambiguous: Affirmative action required (no pre-ticked boxes)

Invalid Consent Examples:

  • ❌ Pre-checked boxes
  • ❌ Bundled consent for unrelated purposes
  • ❌ Consent as condition for service (unless necessary)
  • ❌ Unclear or ambiguous language

Agents MUST allow separate consent for distinct processing purposes.

Example:

{
"consent": {
"service_delivery": {
"granted": true,
"timestamp": "2026-01-15T10:00:00Z",
"required": true
},
"analytics": {
"granted": false,
"timestamp": null,
"required": false
},
"marketing": {
"granted": false,
"timestamp": null,
"required": false
}
}
}

Agents MUST provide mechanism to withdraw consent as easily as it was given (GDPR Art. 7(3)).

Implementation: Preference center, API endpoints, in-app controls.

Agents MUST maintain records of consent:

  • Who gave consent
  • When consent was given
  • What information was provided
  • How consent was obtained
  • Whether consent has been withdrawn

Retention: Consent records must be retained for regulatory compliance even after consent withdrawal.

Agents processing children’s data (under 16, or lower national threshold) MUST:

  • Verify age before processing
  • Obtain parental consent for children under threshold (GDPR Art. 8)
  • Implement age-appropriate privacy notices

Implementation: Age gates, parental verification flows, simplified privacy notices.


Agents transferring personal data outside the EEA MUST use approved mechanisms:

  1. Adequacy Decision (GDPR Art. 45)
  2. Standard Contractual Clauses (SCCs) (GDPR Art. 46(2)(c))
  3. Binding Corporate Rules (BCRs) (GDPR Art. 46(2)(b))
  4. Certification Mechanisms (GDPR Art. 46(2)(f))

Agents MUST document all cross-border data transfers:

  • Recipient country
  • Transfer mechanism (adequacy, SCCs, etc.)
  • Data categories transferred
  • Purpose of transfer
  • Security measures

Implementation: Data flow mapping, transfer impact assessments.

Agents MUST assess destination country laws and supplement transfer mechanisms if necessary (CJEU C-311/18 Schrems II).

Assessment Required:

  • Government access laws
  • National security surveillance
  • Lack of legal remedies
  • Risk to data subjects

[REQ-PRIV-030] Layer 2 (Memory) Privacy Controls

Section titled “[REQ-PRIV-030] Layer 2 (Memory) Privacy Controls”

Agents MUST implement memory-layer privacy controls:

  • Personal data tagging and classification
  • Automated retention policies
  • Secure deletion mechanisms
  • Access logging

[REQ-PRIV-031] Layer 4 (Reasoning) Privacy Controls

Section titled “[REQ-PRIV-031] Layer 4 (Reasoning) Privacy Controls”

Agents MUST implement reasoning-layer privacy controls:

  • Minimize personal data in prompts
  • Avoid leaking personal data in model outputs
  • Differential privacy for model training
  • Federated learning where appropriate

[REQ-PRIV-032] Layer 5 (Persona) Privacy Declarations

Section titled “[REQ-PRIV-032] Layer 5 (Persona) Privacy Declarations”

Agents MUST declare privacy practices in Persona manifest:

{
"persona": {
"id": "agent-001",
"privacy": {
"collects_personal_data": true,
"data_categories": ["contact", "preferences"],
"purposes": ["service_delivery", "analytics"],
"legal_basis": ["consent", "contract"],
"retention_period": "5 years",
"processors": ["processor-a"],
"transfers": [
{
"country": "US",
"mechanism": "standard_contractual_clauses"
}
],
"rights_contact": "privacy@example.com"
}
}
}

Agents MUST implement automated breach detection for:

  • Unauthorized data access
  • Data exfiltration
  • Accidental data exposure
  • Ransomware/encryption attacks

Implementation: Anomaly detection, access pattern monitoring, DLP controls.

[REQ-PRIV-034] Breach Notification (72 Hours)

Section titled “[REQ-PRIV-034] Breach Notification (72 Hours)”

Agents MUST facilitate breach notification to supervisory authority within 72 hours of awareness (GDPR Art. 33).

Required Information:

  • Nature of breach
  • Data categories and approximate number of records affected
  • Likely consequences
  • Measures taken or proposed
  • DPO contact details

Agents MUST facilitate notification to affected data subjects when breach likely to result in high risk (GDPR Art. 34).

Implementation: Automated notification workflows, email/SMS templates, incident management system integration.


Agents claiming GDPR compliance MUST implement all normative requirements ([REQ-PRIV-001] through [REQ-PRIV-035]).

Agents SHOULD undergo regular privacy audits:

  • Data Protection Impact Assessment (DPIA) for high-risk processing
  • Internal privacy audits (annual)
  • Third-party privacy certifications (ISO 27701, etc.)

Conformance testing MUST verify:

  • Data subject rights workflows (access, deletion, portability, etc.)
  • Consent management flows
  • Breach notification procedures
  • Cross-border transfer controls
  • Retention policy enforcement

Agents MUST maintain Records of Processing Activities (RoPA) per GDPR Art. 30:

  • Controller/processor identity
  • Processing purposes
  • Data subject categories
  • Personal data categories
  • Recipient categories
  • Cross-border transfers
  • Retention periods
  • Security measures

Appendix A: Privacy Breach Notification Flow (Informative)

Section titled “Appendix A: Privacy Breach Notification Flow (Informative)”
DataSubjectSupervisoryAuthorityDPOBreachDetectionAgentDataSubjectSupervisoryAuthorityDPOBreachDetectionAgentalt[High risk to data subjects][Low risk]Anomaly detectedConfirm breachNotify breachAssess risk & impactNotify within 72hDirect notificationNotify within 72hDocument decision not to notify subjectsRemediation measuresImplement fixes

Appendix B: Data Subject Rights API (Informative)

Section titled “Appendix B: Data Subject Rights API (Informative)”

Example RESTful API for data subject rights:

POST /privacy/access-request
POST /privacy/rectification
POST /privacy/erasure
POST /privacy/portability-export
POST /privacy/restriction
POST /privacy/objection
GET /privacy/request/{id}/status
DELETE /privacy/data/{subject-id}

Section titled “Appendix C: Consent Management Schema (Informative)”
{
"$schema": "http://json-schema.org/draft-07/schema#",
"type": "object",
"properties": {
"subject_id": { "type": "string" },
"consents": {
"type": "array",
"items": {
"type": "object",
"properties": {
"purpose": { "type": "string" },
"granted": { "type": "boolean" },
"timestamp": { "type": "string", "format": "date-time" },
"method": { "enum": ["opt-in", "opt-out"] },
"required": { "type": "boolean" },
"withdrawn_at": { "type": "string", "format": "date-time" }
},
"required": ["purpose", "granted", "timestamp"]
}
}
}
}


License: This specification is licensed under CC-BY-4.0.
Copyright: © 2026 ARAL Standard Contributors