Article 50 GuideEU AI Act

EU AI Act Article 50 Transparency, Labeling, and User Disclosures

Article 50 is not one generic notice requirement. It separates provider duties for AI systems that interact with people or generate synthetic content from deployer duties for emotion recognition, biometric categorisation, deepfakes, and certain public-interest text.

Use this page to decide which disclosure belongs in the product interface, content workflow, model-output pipeline, workplace or public-facing notice, and which high-risk AI instructions for use must still be handled separately.

Author
Sorena AI
Published
May 9, 2026
Updated
May 9, 2026
Sections
6

Structured answer sets in this page tree.

Primary sources
3

Cited legal and guidance references.

Publication metadata
Sorena AI
Published May 9, 2026
Updated May 9, 2026
Overview

The EU AI Act's transparency rules for labeling and user disclosures are concentrated in Article 50, with related high-risk AI transparency requirements in Article 13 and deployer operating duties in Article 26. A useful implementation record should identify the exact AI system, the actor with the duty, the exposed audience, the timing of the notice, the channel where the notice appears, and the evidence that the notice or marking was actually delivered.

Section 1

Article 50 duties at a glance

Article 50 creates several distinct transparency duties. Providers must design direct-interaction AI systems so natural persons are informed that they are interacting with an AI system, unless that is obvious in context. Providers of AI systems, including general-purpose AI systems, that generate synthetic audio, image, video, or text content must ensure outputs are machine-readable and detectable as artificially generated or manipulated, subject to the Article 50 exceptions.

Deployers carry separate duties where their use exposes people to emotion recognition or biometric categorisation systems, creates or manipulates image, audio, or video content constituting a deepfake, or publishes AI-generated or manipulated text to inform the public on matters of public interest. Article 50 also requires the information in paragraphs 1 to 4 to be clear, distinguishable, accessible, and given no later than the first interaction or exposure.

  • Provider interaction notice: applies to AI systems intended to interact directly with natural persons, unless the interaction is obvious to a reasonably well-informed, observant, and circumspect person in context.
  • Provider synthetic-content marking: applies to AI systems, including GPAI-based systems, that generate synthetic audio, image, video, or text content, with machine-readable and detectable marking as far as technically feasible.
  • Deployer exposure notice: applies when natural persons are exposed to an emotion recognition system or biometric categorisation system.
  • Deployer content disclosure: applies to deepfake image, audio, or video content and to certain AI-generated or manipulated text published to inform the public on matters of public interest.
  • Timing and accessibility: the notice must reach the natural person concerned by the first interaction or exposure and conform to applicable accessibility requirements.
Section 2

Interaction with users: when people must know they are dealing with AI

For a chatbot, voice assistant, agent, embedded support flow, or similar interface, the first question is whether the AI system is intended to interact directly with natural persons. If yes, the provider should treat the disclosure as a product design requirement, not a policy document hidden away from the user journey.

The notice can be unnecessary only when the AI nature is obvious from the viewpoint and context described in Article 50. That exception should be used narrowly in product review: record the interface, the user group, the context of use, and why the AI interaction would be obvious before relying on it.

  • Place the notice where the person starts the interaction, such as the chat entry point, voice prompt, intake form, or generated-response surface.
  • Use plain wording that identifies the interaction as involving an AI system without overstating capability, autonomy, accuracy, or human review.
  • Keep localization and accessibility in scope when the interface is available across languages, channels, or assistive technologies.
  • Retain screenshots, release notes, UI copy approvals, localization records, and accessibility checks as evidence that the notice was available by first interaction.
  • If the feature is also high-risk, keep this Article 50 user-facing notice separate from Article 13 instructions for use supplied to deployers.
Section 3

Synthetic content, deepfakes, and public-interest text

Article 50 separates provider-side technical marking from deployer-side public disclosure. Providers of systems that generate synthetic audio, image, video, or text content must ensure outputs are marked in a machine-readable format and detectable as artificially generated or manipulated, as far as technically feasible and taking account of content type, implementation cost, and the state of the art.

Deployers have a different obligation when they generate or manipulate image, audio, or video content constituting a deepfake: disclose that the content has been artificially generated or manipulated. For evidently artistic, creative, satirical, fictional, or analogous works, the obligation is limited to disclosure of the existence of generated or manipulated content in an appropriate way that does not hamper the display or enjoyment of the work. For AI-generated or manipulated text published to inform the public on matters of public interest, deployers must disclose the AI generation or manipulation unless Article 50's law-enforcement or human-review/editorial-responsibility exceptions apply.

  • Provider evidence should cover watermarking or other marking design, detectability testing, known limitations by content type, interoperability assumptions, and release controls.
  • Deployer evidence for deepfakes should identify the content asset, audience, publication channel, disclosure wording, placement, timing, and any creative-work rationale.
  • For public-interest text, record whether the text was published to inform the public, whether human review or editorial control occurred, and who held editorial responsibility.
  • Do not treat provider machine-readable marking as a substitute for a deployer disclosure where Article 50(4) applies to a publication or content release.
  • Monitor the AI Office code-of-practice process because it is intended to support practical implementation of marking and labeling obligations for AI-generated content.
Section 4

Emotion recognition and biometric categorisation notices

Deployers of emotion recognition systems or biometric categorisation systems must inform the natural persons exposed to the operation of the system. That disclosure duty is separate from data-protection duties: Article 50 expressly states that personal data must be processed under the applicable EU data-protection instruments.

The notice record should therefore answer two questions at once. First, did the exposed person receive clear information about the operation of the system by the time of exposure? Second, did the privacy team assess the lawful basis, data categories, safeguards, and documentation required by the applicable data-protection regime?

  • Define the exposure point: camera zone, kiosk, workplace process, educational process, access-control process, customer journey, or other system context.
  • Separate Article 50 notice wording from privacy-notice wording, then check that both are consistent and available before exposure.
  • Record whether any law-enforcement exception is being relied on and require legal approval before omitting a notice on that basis.
  • Keep signage, interface notices, DPIA or privacy-assessment references, vendor specifications, and deployment maps with the Article 50 evidence record.
  • Screen the use case against Article 5 prohibitions and Annex III high-risk categories before treating Article 50 notice as the only control.
Section 5

Relationship to high-risk AI instructions for use

Article 50 transparency duties do not replace Chapter III high-risk AI requirements. Article 50(6) says paragraphs 1 to 4 do not affect Chapter III obligations, while Article 13 requires high-risk AI systems to be sufficiently transparent for deployers and accompanied by concise, complete, correct, clear, relevant, accessible, and comprehensible instructions for use.

That means the same AI product can need two different transparency artifacts: a user-facing Article 50 notice or label, and a deployer-facing Article 13 instruction set. Article 26 then requires deployers of high-risk AI systems to use the system in accordance with the instructions for use, assign competent human oversight, monitor operation based on those instructions, and keep logs where the logs are under their control.

  • Use Article 50 for notices and labels aimed at natural persons or published content audiences.
  • Use Article 13 for provider instructions that let deployers understand intended purpose, capabilities, limitations, output interpretation, human oversight, maintenance, logging, and relevant risks.
  • Use Article 26 to check whether the deployer has operating procedures, trained oversight, monitoring, incident escalation, and log retention aligned with the instructions for use.
  • For high-risk systems that also generate content or interact directly with people, link the two evidence sets but do not merge them into a single generic transparency checklist.
  • When a deployer changes intended purpose or makes substantial modifications, review whether responsibilities along the AI value chain have changed before reusing old notices.
Section 6

Implementation evidence checklist

Use this checklist for product release, content publication, procurement onboarding, or deployment approval when Article 50 may apply. The goal is to prove that the correct actor delivered the correct notice, label, or technical marking to the correct audience at the correct moment.

Do not use this checklist as a substitute for high-risk classification, prohibited-practice screening, privacy review, or sector-specific disclosure rules. Article 50 transparency can sit beside those duties, and Article 50(6) preserves other transparency obligations under Union or national law.

Does an EU AI Act Article 50 chatbot notice replace high-risk AI instructions for use?

No. Article 50 covers user-facing transparency for direct interaction with an AI system. If the system is high-risk, Article 13 instructions for use and Article 26 deployer operating duties still need their own evidence and controls.

Who discloses AI-generated deepfake content under EU AI Act Article 50?

The deployer of the AI system that generates or manipulates image, audio, or video content constituting a deepfake must disclose that the content was artificially generated or manipulated, subject to the Article 50 exceptions and special treatment for evidently artistic, creative, satirical, fictional, or analogous works.

What should providers keep as evidence for AI-generated content marking under EU AI Act Article 50?

Providers should keep the marking design, machine-readable format decision, detectability testing, content-type limitations, robustness and interoperability assumptions, release approvals, and records showing why any Article 50 exception was or was not used.

What should deployers tell people exposed to emotion recognition or biometric categorisation systems?

Deployers should inform the natural persons exposed to the operation of the emotion recognition or biometric categorisation system by the time of exposure, and keep the Article 50 notice aligned with applicable EU personal-data compliance records.

  • Classify the trigger: direct interaction, synthetic content generation, emotion recognition exposure, biometric categorisation exposure, deepfake content, or public-interest text publication.
  • Name the actor with the duty: provider for direct-interaction design and synthetic-output marking; deployer for exposed-person notices, deepfake disclosure, and public-interest text disclosure.
  • Define the audience and timing: natural person at first interaction or exposure, content viewer at publication, or deployer receiving high-risk instructions for use.
  • Approve the exact disclosure text, placement, language coverage, accessibility treatment, and exception analysis.
  • Save evidence: UI screenshots, marking specifications, detectability tests, publication records, signage or notice copies, editorial review records, DPIA references, Article 13 instructions, Article 26 operating procedures, and change approvals.
  • Set review triggers for new modalities, new publication channels, new user groups, supplier model changes, high-risk classification changes, and changes to the AI Office code-of-practice or Commission guidance.
Primary sources

References and citations

digital-strategy.ec.europa.eu
Referenced sections
  • Commission overview identifies emotion recognition and biometric categorisation in the AI Act risk framework, including prohibited and high-risk examples.
Related guides

Explore more topics

Are industry AI use cases high-risk under EU AI Act Annex III?
FAQ answer on when an industry AI use case falls under EU AI Act Annex III, how Article 6 classification works, when Article 6(3) can support a non-high-risk conclusion, and what evidence providers should keep.
EU AI Act AI System Classification Edge Cases FAQ
Answers for EU AI Act edge cases: AI system definition, inference versus simple rules, GPAI models, embedded products, territorial scope, roles, and classification evidence.
EU AI Act Applicability and Roles: Scope, Actor Map, and Evidence
Determine whether the EU AI Act applies to an AI system or GPAI model, map provider, deployer, importer, distributor, and product manufacturer roles, and record evidence for classification.
EU AI Act applicability test: scope, role, and risk classification
Stepwise EU AI Act applicability test for AI-system status, exclusions, territorial scope, operator role, prohibited uses, high-risk systems, GPAI models, transparency duties, and evidence records.
EU AI Act Article 5 Prohibited AI Practices Screening Guide
Screen AI systems against the EU AI Act Article 5 prohibitions, including manipulation, exploitation, social scoring, biometric and law-enforcement exceptions.
EU AI Act Article 50 transparency disclosures FAQ
Article 50 FAQ for EU AI Act transparency duties covering chatbot notices, synthetic content marking, biometric and emotion notices, deepfakes, public-interest text, timing, accessibility, and exceptions.
EU AI Act Article 73 serious incident FAQ
FAQ on EU AI Act serious incident handling for high-risk AI systems, including Article 73 reporting, deployer escalation, corrective action, and GPAI systemic-risk distinctions.
EU AI Act Compliance Checklist by Risk Class
A practical EU AI Act checklist for classifying AI systems, assigning operator roles, screening prohibited practices, and collecting evidence for high-risk, GPAI, transparency, monitoring, and incident duties.
EU AI Act Compliance Program: roles, high-risk evidence, GPAI and incidents
Build an EU AI Act compliance program around provider, deployer, importer, distributor, high-risk, GPAI, transparency, monitoring, and incident evidence duties.
EU AI Act conformity assessment and notified bodies for high-risk AI
Grounded guide to EU AI Act high-risk AI conformity assessment routes, provider evidence, EU declaration of conformity, CE marking, and notified body involvement.
EU AI Act deadlines and compliance calendar | Article 113 dates
source-linked EU AI Act compliance calendar for Article 113 staged application dates, Article 111 transitions, GPAI, prohibited practices, AI literacy, and high-risk AI planning.
EU AI Act FAQ: scope, roles, high-risk AI, GPAI, FRIA, and dates
Grounded EU AI Act FAQ covering scope, provider and deployer roles, prohibited practices, high-risk classification, GPAI duties, transparency notices, FRIAs, EU database registration, serious incidents, and staged application dates.
EU AI Act FRIA FAQ: Article 27 Scope, Contents, and Notification
Source-grounded FAQ on when Article 27 requires a fundamental rights impact assessment, which deployers are covered, what the FRIA must contain, and how it relates to DPIAs and registration.
EU AI Act FRIA for high-risk AI systems: Article 27 scope and evidence
Source-grounded guide to EU AI Act Article 27 fundamental rights impact assessments: who must run a FRIA, Article 6(2) triggers, Annex III carveouts, DPIA overlap, notification, and registration evidence.
EU AI Act GPAI and Systemic-Risk Duties: Article 53 and 55 FAQ
FAQ on EU AI Act duties for general-purpose AI model providers, including Article 53 documentation, copyright and training-summary duties, Article 55 systemic-risk duties, serious incidents, cybersecurity, and staged enforcement.
EU AI Act GPAI evidence pack checklist for Article 53 and 55
Build a source-grounded evidence pack for EU AI Act GPAI model obligations: technical documentation, downstream information, copyright policy, training-content summary, and systemic-risk records where applicable.
EU AI Act GPAI Provider Obligations: Articles 53 and 55
Grounded guide to EU AI Act duties for general-purpose AI model providers: Article 53 documentation, copyright policy, training-content summary, downstream information, and Article 55 systemic-risk controls.
EU AI Act High-Risk AI Requirements: Articles 8-16 and 26
Map the EU AI Act requirements for high-risk AI systems: risk management, data governance, technical documentation, logs, transparency, human oversight, accuracy, robustness, cybersecurity, and deployer duties.
EU AI Act high-risk AI use cases by industry | Article 6 and Annex III guide
Industry-by-industry guide to EU AI Act high-risk classification under Article 6, Annex III, Annex I product safety routes, exclusions, and provider/deployer boundaries.
EU AI Act high-risk conformity assessment route selector
Select the EU AI Act Article 43 conformity assessment route for a high-risk AI system, including Annex I product legislation, Annex III categories, notified body triggers, standards, declaration, CE marking, registration, and evidence.
EU AI Act high-risk requirements checklist: Articles 8-15
Checklist for EU AI Act high-risk AI system requirements in Articles 8-15: risk management, data governance, documentation, logs, transparency, human oversight, accuracy, robustness, and cybersecurity.
EU AI Act penalties and fines: Article 99 tiers and GPAI exposure
EU AI Act penalties explained: Article 99 fine tiers, prohibited-practice exposure, incorrect information, SME caps, Member State rules, and GPAI model fines.
EU AI Act post-market monitoring and serious incident reporting
Grounded guide to EU AI Act Articles 72 and 73 for high-risk AI: monitoring plans, serious incident reporting, deployer escalation, corrective action, and GPAI distinctions.
EU AI Act post-market monitoring FAQ for high-risk AI systems
Answer to how providers and deployers should handle EU AI Act post-market monitoring for high-risk AI systems under Article 72, with serious-incident, log, corrective-action, and lifecycle-change triggers.
EU AI Act provider vs deployer role boundaries: Article 3 and Article 25 FAQ
FAQ on EU AI Act provider, deployer, operator, importer, distributor, authorised representative, product manufacturer, downstream provider, and GPAI model provider boundaries.
EU AI Act risk classification intake workflow
A grounded intake structure for classifying EU AI Act scope, prohibited practices, high-risk routes, Annex III use cases, GPAI model status, roles, and reassessment triggers.
EU AI Act serious incident reporting triage workflow: Article 73 and Article 55
Triage EU AI Act serious incidents by definition, actor, reporting route, deadline, deployer escalation, corrective action, and separate GPAI systemic-risk reporting.
EU AI Act Technical Documentation and Provider Evidence Templates
Build AI Act evidence templates for high-risk AI providers: Article 11 technical documentation, Annex IV fields, quality management, conformity, CE marking, registration, logs, and post-market monitoring.
EU AI Act technical documentation FAQ | Article 11 and Annex IV
What Article 11 and Annex IV require in high-risk AI technical documentation: system identity, intended purpose, architecture, data, testing, oversight, cybersecurity, conformity, and post-market monitoring.
EU AI Act Timeline and Phasing Roadmap: practical obligations and evidence guide
Practical EU AI Act guide to Timeline and Phasing Roadmap: scope, owners, evidence, edge cases, checklist steps, and external source-linked citations.
EU AI Act vs ISO/IEC 42001: legal duties, controls, and evidence limits
Compare the EU AI Act and ISO/IEC 42001 across legal status, risk classification, high-risk AI, GPAI, transparency, conformity, evidence, and assurance limits.
EU AI Act vs NIST AI RMF: legal duties, risk controls, and evidence boundaries
Compare the binding EU AI Act with the voluntary NIST AI RMF, including role classification, high-risk duties, GPAI, transparency, conformity evidence, and reuse limits.
FAQ: EU AI Act conformity assessment procedures and notified body selection
source-linked FAQ on EU AI Act Article 43 conformity assessment routes, Annex VI internal control, Annex VII notified-body review, CE marking, declarations, and registration.