Skip to main content

​
AIGO — EU AI Act Transparency Mapping

​
1. Document Purpose

This document provides the AIGO mapping for the transparency obligations established by Article 50 of Regulation (EU) 2024/1689, as amended by subsequent Union legislation including Regulation (EU) 2026/1744. The mapping translates the Article 50 framework into AIGO governance mechanisms covering:
  • applicability;
  • actor identification;
  • AI interaction disclosure;
  • machine-readable marking;
  • AI-generated and manipulated content;
  • deepfakes;
  • AI-generated text concerning matters of public interest;
  • emotion-recognition systems;
  • biometric categorisation systems;
  • human review and editorial control;
  • technical safeguards;
  • transparency evidence;
  • monitoring;
  • incidents;
  • change management;
  • assurance;
  • management review; and
  • continual improvement.
The document is an operational governance mapping. It is not legal advice, a legal opinion, or a declaration of compliance. The European Commission adopted its final Article 50 transparency guidelines on 20 July 2026. The guidelines state that Article 50 transparency obligations apply from 2 August 2026 and provide practical guidance on scope, exemptions, implementation, and demonstration of compliance.

​
2. Mapping Information

​
3. Source Hierarchy

The primary source is Article 50 of the AI Act as contained in: Regulation (EU) 2024/1689 as amended by applicable subsequent Union legislation. The current legal text on EUR-Lex remains authoritative.

​
3.2 Official Implementation Material

Supporting material includes:
  • European Commission Article 50 guidelines;
  • European Commission transparency guidance;
  • Code of Practice on Transparency of AI-Generated Content;
  • official FAQs;
  • other official implementation material.
The Commission states that the Article 50 guidelines clarify scope, concepts, exemptions, practical application, and ways of demonstrating compliance.

​
3.3 AIGO Mapping

AIGO translates the legal requirement into operational governance structures. AIGO implementation mechanisms must not be described as automatically equivalent to statutory compliance.

​
4. Transparency Governance Principle

AIGO should treat Article 50 transparency as a controlled lifecycle obligation. The recommended governance sequence is:
Transparency should be considered during design and configuration rather than added only after deployment.

​
5. Current Applicability

The European Commission states that Article 50 transparency obligations apply from: 2 August 2026. The current AIGO mapping should therefore treat Article 50 as a current operational mapping area as of the date of this document. Certain marking obligations have a transitional treatment for some systems placed on the market before 2 August 2026. The Commission identifies a transition until 2 December 2026 for specified generative-AI systems already on the market before the application date. The detailed timeline is maintained centrally in: 11-AIGO-EU-AI-Act-Applicability-and-Timeline-v0.1.md

​
6. AIGO Transparency Control

Control Name: EU AI Act Transparency Control Objective: Ensure that applicable AI systems and AI-generated or manipulated content are transparently identified in accordance with the applicable Article 50 requirements. Control Owner: AI System Owner, with AI Governance, Legal/Compliance, Product, and Technical ownership as applicable. Control Frequency:
  • before deployment;
  • before material changes;
  • before enabling new content-generation capabilities;
  • periodically during operation;
  • when official guidance changes;
  • when applicable system or deployment conditions change.
Required Evidence:
  • transparency applicability assessment;
  • system configuration;
  • notices;
  • content marking implementation;
  • testing;
  • monitoring;
  • exceptions;
  • review;
  • approval.

​
7. Article 50 Transparency Architecture

Article 50 should be mapped as four primary operational areas:
The Commission’s current summary identifies four principal transparency situations:
  • individuals interacting directly with AI;
  • exposure to AI-generated or manipulated content;
  • exposure to emotion recognition or biometric categorisation;
  • AI-generated or manipulated text concerning matters of public interest without human review or editorial control.

​
8. Article 50(1) — Direct Interaction with AI

Providers of AI systems intended to interact directly with natural persons must design and develop the systems so that the persons are informed that they are interacting with an AI system, subject to the legal conditions and exceptions.

​
8.2 AIGO Mapping

AIGO Components:
  • AI System;
  • Governance;
  • Control;
  • Assessment;
  • Evidence;
  • Monitoring;
  • Assurance.
Relationship: DIRECT

​
8.3 AIGO Control

AI Interaction Disclosure Control The control should require:
  • identification of direct AI interaction;
  • applicability screening;
  • disclosure method;
  • timing of disclosure;
  • visibility;
  • accessibility;
  • user testing;
  • change control.

​
8.4 Evidence

Potential evidence:
  • interface design;
  • disclosure configuration;
  • product requirements;
  • screenshots;
  • test results;
  • accessibility assessment;
  • user acceptance testing;
  • monitoring.
The Commission’s guidelines clarify the scope of directly interactive AI systems and provide practical examples of systems that fall within or outside the obligation.

​
9. Direct Interaction Disclosure Principle

The disclosure should be designed so that an individual can reasonably understand that they are interacting with AI. AIGO should evaluate:
  • prominence;
  • timing;
  • clarity;
  • accessibility;
  • persistence where appropriate;
  • consistency across interaction channels;
  • changes after system updates.
A technical field containing disclosure text is not automatically sufficient if users do not reasonably encounter it.

​
10. Interaction Disclosure Evidence

Evidence may include:
Where the system operates through multiple channels, evidence should cover the applicable channels.

​
11. Interaction Disclosure Monitoring

Monitoring may include:
  • disclosure availability;
  • disclosure failures;
  • user complaints;
  • interface changes;
  • language changes;
  • channel-specific failures;
  • unauthorized interface modifications.
Material failures should trigger incident or change governance.

​
12. Interaction Disclosure Change Governance

The transparency assessment should be repeated after:
  • interface redesign;
  • model-provider change;
  • user journey change;
  • deployment-channel change;
  • new language;
  • new target population;
  • new modality;
  • new embedded-AI capability.

​
13. Article 50(2) — Machine-Readable Marking of AI-Generated or Manipulated Content

Providers of AI systems that generate synthetic audio, image, video, or text content must ensure that the outputs are marked in a machine-readable format and detectable as artificially generated or manipulated, subject to the legal conditions and applicable exceptions. The Commission’s 2026 transparency framework identifies Article 50(2) as one of the provisions supported by the Code of Practice on Transparency of AI-Generated Content.

​
13.2 AIGO Mapping

AIGO Components:
  • AI System;
  • Control;
  • Evidence;
  • Monitoring;
  • Assurance;
  • Change.
Relationship: DIRECT / CRITICAL

​
14. AI-Generated Content Marking Control

Control Name: Machine-Readable AI Content Marking Control Objective: Ensure that applicable AI-generated or manipulated content is marked in a technically appropriate machine-readable way that enables reliable detection. The implementation should address:
  • output type;
  • marking method;
  • persistence;
  • interoperability;
  • detection;
  • robustness;
  • modification;
  • downstream processing;
  • testing.

​
15. Marking Governance

The organization should establish:
  • supported marking standard or technique;
  • implementation owner;
  • technical requirements;
  • testing;
  • monitoring;
  • change management;
  • exception handling;
  • evidence retention.
Where the organization uses the Commission-approved Code of Practice as its implementation route, the adoption decision should itself be documented.

​
16. Code of Practice Relationship

The European Commission and AI Board concluded in July 2026 that the Code of Practice on Transparency of AI-Generated Content adequately covers the Article 50(2), (4), and (5) obligations. The Commission describes the code as a voluntary instrument that facilitates compliance. Adherence is not, however, conclusive evidence of compliance. AIGO should therefore represent the relationship as:
The code should not be treated as an independent legal source equivalent to Article 50.

​
17. Alternative Compliance Measures

The Commission’s final guidelines state that providers and deployers of generative AI systems that do not adhere to the code must demonstrate compliance through other appropriate and equivalently adequate means. AIGO should therefore support:
The chosen approach should be documented.

​
18. Marking Integrity

AIGO should assess whether machine-readable marking remains available after foreseeable downstream processing. Relevant considerations may include:
  • format conversion;
  • compression;
  • editing;
  • distribution;
  • publication;
  • re-export;
  • platform ingestion.
The technical implementation should be tested rather than assumed to be robust.

​
19. Marking Evidence

Potential evidence includes:
  • marking specification;
  • implementation architecture;
  • technical configuration;
  • test results;
  • interoperability testing;
  • red-team testing;
  • platform compatibility testing;
  • change records;
  • monitoring.

​
20. Marking Exception and Transition

The Commission’s current transparency material states that certain generative-AI systems placed on the market before 2 August 2026 benefit from a transition for the marking obligation until 2 December 2026. AIGO should therefore distinguish:
A transitional status should not be represented as permanent exemption.

​
21. Article 50(3) — Deepfake Disclosure

Deployers of AI systems that generate or manipulate image, audio, or video content constituting a deepfake must disclose that the content has been artificially generated or manipulated, subject to applicable legal conditions and exceptions.

​
21.2 AIGO Mapping

AIGO Components:
  • Control;
  • Transparency;
  • Evidence;
  • Monitoring;
  • Incident.
Relationship: DIRECT

​
22. Deepfake Disclosure Control

Control Name: Deepfake Disclosure Control The control should require:
  • detection of deepfake content;
  • applicable disclosure method;
  • publication workflow;
  • placement;
  • visibility;
  • record keeping;
  • testing.
Where technical detection is used, the detection process should itself be tested.

​
23. Deepfake Disclosure Evidence

Evidence may include:
  • content classification;
  • detection output;
  • disclosure configuration;
  • publication record;
  • user-facing label;
  • review record;
  • technical testing.

​
24. Deepfake Human Review

Where human review or editorial control affects the applicability of an obligation, AIGO should record:
  • reviewer;
  • role;
  • review process;
  • decision;
  • date;
  • evidence.
A system should not simply record “human reviewed” without specifying what review occurred.

​
25. Deepfake Monitoring

Monitoring should consider:
  • unlabeled deepfakes;
  • incorrect labels;
  • disclosure failures;
  • content pipeline failures;
  • platform transformation;
  • user complaints.
Material failures should trigger Incident or Improvement management.

​
26. Article 50(4) — AI-Generated or Manipulated Text on Matters of Public Interest

Deployers of AI systems that generate or manipulate text published to inform the public on matters of public interest must disclose that the text was artificially generated or manipulated, where the statutory conditions apply and where the content has not undergone appropriate human review or editorial control. The Commission’s current guidance specifically addresses AI-generated text on matters of public interest and the concepts of human review and editorial control.

​
26.2 AIGO Mapping

AIGO Components:
  • Governance;
  • Transparency;
  • Human Oversight;
  • Control;
  • Evidence;
  • Monitoring;
  • Assurance.
Relationship: DIRECT / CONDITIONAL

​
27. Public-Interest Text Control

Control Name: Public-Interest AI Text Disclosure Control The control should require:
  • identification of public-interest publication;
  • AI-generation determination;
  • human review determination;
  • editorial-control determination;
  • disclosure applicability;
  • publication decision;
  • evidence.

​
28. Human Review and Editorial Control

AIGO should define these concepts operationally while preserving the legal meaning. The record should identify:
  • reviewer;
  • reviewer competence;
  • scope of review;
  • factual review;
  • editorial review;
  • corrections;
  • approval;
  • publication.
A nominal human click or automated acceptance should not automatically be represented as meaningful human review.

​
29. Public-Interest Text Evidence

Potential evidence includes:
  • content-generation records;
  • editorial workflow;
  • reviewer identity;
  • review record;
  • publication record;
  • disclosure label;
  • version history;
  • correction history.

​
30. Public-Interest Monitoring

Monitoring should address:
  • unlabeled content;
  • incorrect classification;
  • missing human review;
  • disclosure failures;
  • unauthorized publishing;
  • complaints;
  • correction events.

​
31. Article 50(5) — Emotion Recognition and Biometric Categorisation

Deployers of AI systems that perform emotion recognition or biometric categorisation must inform exposed individuals in the circumstances covered by Article 50.

​
31.2 AIGO Mapping

AIGO Components:
  • Governance;
  • Control;
  • Privacy;
  • Human Oversight;
  • Evidence;
  • Monitoring;
  • Risk.
Relationship: DIRECT / CONDITIONAL

​
32. Emotion Recognition Disclosure Control

Control Name: Emotion Recognition Transparency Control The control should require:
  • system identification;
  • applicability analysis;
  • disclosure;
  • affected-person visibility;
  • timing;
  • accessibility;
  • evidence.
Article 50 transparency should be assessed separately from Article 5, because some emotion-recognition applications may be prohibited while other applications may instead be subject to transparency requirements or other rules.

​
33. Biometric Categorisation Disclosure Control

The control should distinguish:
  • biometric identification;
  • biometric verification;
  • biometric categorisation;
  • biometric data processing;
  • emotion recognition.
The organization should not treat all biometric technologies as a single regulatory category.

​
34. Article 5 Relationship

Article 50(5) may intersect with Article 5. For example:
A system should not use Article 50 compliance as a justification for an otherwise prohibited use.

​
35. Transparency Applicability Matrix

The exact applicability depends on the legal conditions, exceptions, and current guidance.

​
36. Transparency Risk Classification

AIGO should recognize transparency as a distinct governance risk category. Potential risk: TRANSPARENCY_RISK Potential consequences include:
  • deception;
  • uninformed interaction;
  • misinformation;
  • inability to recognize synthetic content;
  • improper biometric exposure;
  • regulatory non-compliance;
  • reputational harm.

​
37. Transparency Risk Assessment

The risk assessment should consider:
  • user exposure;
  • content type;
  • affected-person scale;
  • public-interest significance;
  • system capability;
  • disclosure effectiveness;
  • misuse;
  • technical failures;
  • downstream transformation.
Risk treatment should connect directly to transparency controls.

​
38. Transparency Assessment

An AIGO transparency assessment should determine:
  • whether Article 50 applies;
  • which subsection applies;
  • which actor has the obligation;
  • whether an exception applies;
  • required measure;
  • implementation status;
  • evidence status;
  • monitoring;
  • assurance.

​
39. Transparency Approval

A transparency-related deployment decision may require approval when:
  • applicability is uncertain;
  • the system is high-impact;
  • disclosure is technically complex;
  • the organization uses an alternative compliance mechanism;
  • the content concerns sensitive public-interest matters;
  • material exception interpretation is required.
Approval does not replace any legal requirement.

​
40. Transparency Evidence

Recommended evidence categories include:

​
41. Transparency Evidence Quality

For material obligations, evidence should be:
  • attributable;
  • current;
  • traceable;
  • relevant;
  • complete;
  • protected against unauthorized modification.
The Evidence Coverage Validator should be capable of identifying missing transparency evidence.

​
42. Transparency Monitoring

Monitoring should cover:
  • disclosure availability;
  • marking integrity;
  • detection success;
  • false positives;
  • false negatives;
  • changes;
  • user complaints;
  • regulatory findings.
Where automated detection is used, model/system performance should be monitored.

​
43. Transparency Incident Management

Potential transparency incidents include:
  • missing AI disclosure;
  • missing machine-readable mark;
  • broken content mark;
  • unlabeled deepfake;
  • missing public-interest disclosure;
  • unauthorized emotion-recognition deployment;
  • incorrect biometric-category notification.
Incidents should connect to:
  • affected system;
  • risk;
  • control;
  • evidence;
  • change;
  • improvement.

​
44. Transparency Change Management

A change should trigger Article 50 reassessment where it affects:
  • generation capability;
  • output type;
  • publication process;
  • interaction mode;
  • interface;
  • biometric functionality;
  • emotion-recognition functionality;
  • public-interest publication;
  • marking implementation.

​
45. Transparency Assurance

Assurance may evaluate:
  • applicability;
  • disclosure effectiveness;
  • technical marking;
  • human review;
  • labeling;
  • evidence;
  • monitoring;
  • exception handling.
For machine-readable marking, technical testing may be more important than document review alone.

​
46. Transparency Assurance Methods

Possible assurance methods include:
  • configuration review;
  • code review;
  • functional testing;
  • interoperability testing;
  • adversarial testing;
  • content sampling;
  • user testing;
  • accessibility assessment;
  • independent review.

​
47. Transparency and the Code of Practice

The Code of Practice on Transparency of AI-Generated Content supports compliance with Article 50(2), (4), and (5). The Commission states that it is an EU-wide adequate instrument for providers and deployers of generative AI systems, while also clarifying that adherence is not conclusive evidence of compliance. AIGO should record whether the organization:
The decision should be supported by evidence.

​
48. Transparency and Alternative Measures

Where the organization does not rely on the code, it should document:
  • alternative measures;
  • legal basis;
  • technical basis;
  • equivalence assessment where appropriate;
  • implementation;
  • evidence;
  • assurance.
The Commission’s guidelines expressly contemplate demonstrating compliance through other appropriate means.

​
49. Transparency and General-Purpose AI

GPAI systems may create transparency obligations in addition to the GPAI-specific requirements. AIGO should therefore evaluate:
The GPAI mapping remains separately authoritative for GPAI-specific obligations.

​
50. Transparency and High-Risk AI

High-risk status and Article 50 transparency status should be recorded separately. A system can be:
  • high-risk and subject to Article 50;
  • high-risk without a particular Article 50 obligation;
  • subject to Article 50 without being high-risk;
  • both high-risk and GPAI-related where the legal structure permits;
  • outside Article 50.
AIGO should not infer one category from another.

​
51. Transparency and Article 5

The relationship must be:
A transparency mechanism cannot legalize a prohibited practice.

​
52. Transparency and Fundamental Rights

Transparency may support:
  • autonomy;
  • informed decision-making;
  • freedom from deception;
  • privacy;
  • dignity;
  • accountability.
AIGO should consider fundamental-rights implications when transparency failures create material harm.

​
53. Transparency and Accessibility

Transparency mechanisms should be accessible to intended users. AIGO may assess:
  • language;
  • visual accessibility;
  • assistive technologies;
  • cognitive accessibility;
  • placement;
  • comprehension.
Accessibility requirements should be derived from applicable law and organizational policy rather than invented as universal legal requirements.

​
54. Transparency by Design

Transparency requirements should be considered during:
  • design;
  • product development;
  • procurement;
  • configuration;
  • testing;
  • deployment;
  • change;
  • monitoring.
AIGO should avoid a model where transparency is added only as a final compliance label.

​
55. Transparency by Procurement

Procurement should require providers to disclose:
  • whether the system generates synthetic content;
  • marking approach;
  • interaction disclosure;
  • biometric functions;
  • emotion recognition;
  • deepfake capability;
  • public-interest content functionality;
  • applicable compliance approach;
  • changes affecting transparency.
Supplier statements should be verified where material.

​
56. Transparency by Configuration

The AI System Profile should capture:
  • applicable Article 50 subsection;
  • disclosure method;
  • marking technology;
  • user-facing notification;
  • publication process;
  • compliance approach;
  • owner;
  • evidence.

​
57. Transparency by Lifecycle

​
58. Transparency Exceptions

Exceptions should be determined according to the Regulation and applicable official guidance. AIGO must distinguish:
from:
An internal waiver cannot override Article 50.

​
59. Transitional Status

AIGO should support:
For content-marking obligations, the Commission currently identifies a transition until 2 December 2026 for certain systems placed on the market before 2 August 2026.

​
60. Transparency Review Triggers

Review should be triggered by:
  • new AI capability;
  • new output type;
  • new publication channel;
  • change to marking method;
  • change to user interface;
  • new biometric functionality;
  • emotion-recognition capability;
  • new provider;
  • new deployment role;
  • regulatory amendment;
  • new Commission guidance.

​
61. Transparency Control Matrix

​
62. Transparency Traceability Chain

The minimum chain should be:
Where a failure occurs:

​
63. Transparency Risk and Control Coverage

The Control Coverage Validator should eventually identify:
  • required transparency control;
  • implemented control;
  • assessed control;
  • evidenced control;
  • monitored control;
  • assured control.
Critical transparency gaps should be escalated according to the applicable governance profile.

​
64. Transparency Evidence Coverage

The Evidence Coverage Validator should eventually verify:
  • applicability evidence;
  • implementation evidence;
  • testing evidence;
  • current marking evidence;
  • disclosure evidence;
  • human-review evidence;
  • monitoring evidence;
  • assurance evidence.
A successful technical implementation without current evidence should not automatically be represented as fully evidenced.

​
65. Transparency Mapping Findings

Potential findings include:

​
66. Critical Transparency Gaps

Potential critical gaps include:
  • required AI-interaction disclosure absent;
  • required machine-readable marking absent;
  • required deepfake disclosure absent;
  • required public-interest disclosure absent;
  • required biometric/emotion-recognition notification absent;
  • technical marking present but not detectable;
  • applicable transparency requirement not assessed;
  • transition incorrectly treated as permanent exemption.

​
67. Transparency Quality Criteria

A high-quality Article 50 implementation should be:
  • legally applicable;
  • clearly attributed;
  • technically implemented;
  • user-visible where required;
  • machine-detectable where required;
  • tested;
  • monitored;
  • evidenced;
  • reviewed;
  • maintained through change management.

​
68. Current Official Guidance Baseline

The European Commission’s final Article 50 guidelines were published on 20 July 2026 and specifically address:
  • who falls within the obligations;
  • relevant definitions;
  • exemptions;
  • directly interactive AI systems;
  • synthetic content;
  • deepfakes;
  • AI-generated text on public-interest matters;
  • emotion recognition;
  • biometric categorisation;
  • ways to demonstrate compliance.
AIGO should use this final guidance as its primary official implementation reference for version 0.1.

​
69. Code of Practice Baseline

The Code of Practice on Transparency of AI-Generated Content supports the obligations related to marking and labelling of AI-generated content. The Commission and AI Board assessed it as an adequate tool for Article 50(2), (4), and (5). AIGO should record:
  • whether the code is adopted;
  • scope of adoption;
  • responsible authority;
  • implementation date;
  • evidence;
  • review status.

​
70. Transparency Legal-Source Currency

The mapping should be reviewed when:
  • Article 50 is amended;
  • Regulation (EU) 2026/1744 is further amended;
  • Commission transparency guidelines change;
  • the transparency Code of Practice is updated;
  • relevant standards change;
  • enforcement interpretations materially develop.
The current official guidance should be checked before each mapping release.

​
71. Relationship to Other EU AI Act Mappings

​
72. Relationship to AIGO Schemas

The transparency mapping should use existing schemas rather than introducing a separate Article 50 record type at this stage.

​
73. Relationship to AIGO Templates

Relevant templates include:
  • AI System Registration;
  • AI System Profile;
  • AI Classification;
  • AI Risk Assessment;
  • AI Control Assessment;
  • AI Approval;
  • AI Monitoring;
  • AI Incident;
  • AI Change Management;
  • AI Assurance;
  • AI Management Review;
  • AI Continuous Improvement;
  • AI Evidence Record.
The organization may later create a dedicated transparency assessment template if operational usage demonstrates a need.

​
74. Relationship to AIGO Tools

The transparency mapping should be usable by:
  • Schema Validator;
  • Reference Validator;
  • Traceability Validator;
  • Control Coverage Validator;
  • Evidence Coverage Validator;
  • Framework Consistency Checker;
  • Document Integrity Checker;
  • Repository Health Checker.

​
75. Validation Requirements

The mapping should satisfy:

​
Source Validation

Article 50 and current amendment references resolve.

​
Guidance Validation

Current official Commission guidance is identified.

​
Applicability Validation

Actor, system, content, and context are identified.

​
Control Validation

Applicable transparency controls exist.

​
Evidence Validation

Required evidence is defined.

​
Timeline Validation

2 August 2026 application and applicable transition rules are correctly represented.

​
Code Validation

Where the Code of Practice is used, adoption and scope are documented.

​
Traceability Validation

Article 50 → AIGO control → evidence chain is complete.

​
Consistency Validation

Terminology matches the master mapping and legal source.

​
76. Limitations

This mapping cannot independently determine:
  • whether a particular disclosure satisfies the exact legal standard in a specific case;
  • whether content is legally a deepfake;
  • whether text concerns a matter of public interest;
  • whether a human review qualifies under the legal standard;
  • whether an exception applies;
  • whether technical marking is legally adequate;
  • whether a specific alternative measure is equivalently adequate.
Those determinations may require legal, technical, and contextual assessment.

​
77. Document Control

​
78. Document Status

Document: AIGO — EU AI Act Transparency Mapping Version: 0.1 Status: Draft Working Name: AIGO Full Name: AI Governance Operating Framework Document Identifier: AIGO-MAP-EUAI-004 Document Type: EU AI Act Mapping This document maps Article 50 transparency obligations to the AIGO AI Governance Operating Framework, including direct AI interaction, machine-readable marking, deepfakes, public-interest AI-generated text, emotion recognition, biometric categorisation, evidence, monitoring, assurance, and lifecycle governance. End of Document