{"content_id":"o8aszqjsw0","slug":"eu-ai-act-article-50-chatbot-deepfake-transparency","locale":"en","schema_type":"Article","category":"ai_data","category_name":"AI Data","title":"EU AI Act Article 50 Chatbot and Deepfake Labels","summary":"EU AI Act Article 50 requires chatbot notices, deepfake disclosures, and other measures from August 2, 2026. Machine-readable labels and human-perceivable disclosures are separate requirements, and some obligations for existing systems are subject to a grace period.","sponsorship_disclosure":null,"affiliate_disclosure":null,"commerce_disclosure":null,"author":{"name":"Injoys Editorial Team","url":"https://injoys.com/ko/about"},"key_points":["The transparency obligations under EU AI Act Article 50 generally apply from August 2, 2026.","Providers are responsible for designing notices for AI systems that interact directly with people and for marking generated content in a machine-readable format.","Deployers must disclose the use of AI in deepfakes and in public-interest text subject to disclosure obligations.","The disclosure exception for public-interest text requires substantive human review or editorial control, as well as editorial responsibility.","The obligations under Article 50(2) for systems placed on the market before August 2, 2026, apply from December 2, 2026."],"content_markdown":"Article 50 of the EU AI Act applies from August 2, 2026. Disclosures for chatbots and deepfakes are central. Machine-readable marking of generated content is also required. However, the responsible party and the exceptions and transition periods differ by obligation.\n\nThe dates and figures in this article are based on the Commission’s official guidance from July and August 2026.\n\n## What Does Article 50 Require?\n\nArticle 50 sets out transparency rules requiring disclosure of AI use. Not every obligation applies to every AI system. The applicable provision depends on the system’s function and the purpose of the content.\n\n| Applicable situation | Primary responsible party | Required action |\n|---|---|---|\n| AI that interacts directly with people | Provider | Design it to disclose that the person is interacting with AI |\n| AI that generates audio, images, video, or text | Provider | Ensure machine-readable marking and detectability of outputs |\n| Emotion recognition or biometric categorisation system | Deployer | Inform affected persons that the system is operating |\n| Deepfake or public-interest text subject to disclosure | Deployer | Disclose that the content was generated or manipulated by AI |\n\nThe Commission published interpretive guidelines on July 20, 2026. In an announcement on July 31 of the same year, it reaffirmed that the rules would apply from August 2. This date does not apply uniformly to every provision of the AI Act. [Commission enforcement announcement](https://digital-strategy.ec.europa.eu/en/news/commission-starts-enforcing-ai-act-rules-and-new-transparency-requirements-2-august)\n\n## Comparing the Responsibilities of Providers and Deployers\n\nResponsibility is divided according to the role actually performed, rather than the company’s industry. A provider is an entity that develops an AI system or has one developed. It places the system on the market or puts it into service under its own name or trademark. A deployer is an entity that uses an AI system under its authority.\n\n| Category | Provider | Deployer |\n|---|---|---|\n| Statutory term | provider | deployer |\n| Typical role | Provides an AI system under its own name | Uses AI to produce content or perform work |\n| Core obligation | Design that enables disclosure and technical marking | Disclosure appropriate to the context of use and exposure |\n| Basis for determination | Responsibility for development, market placement, or putting into service | Authority over the use of the system |\n\nIn this article, “deployer” means deployer. It is a different concept from a distributor involved in sales and distribution. A single organisation may act as both provider and deployer. The definitions are based on [Article 3 of the AI Act](https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-3).\n\nBusinesses outside the EU may also fall within its scope. This includes cases where an AI system is provided in the EU market. It also applies where the system’s output is used in the EU. Being based in South Korea does not, by itself, exclude a business.\n\nDeployer obligations do not apply to purely personal, non-professional use. Business or professional activities are distinct from this exception. Article 50 is not exempted merely because a system is free or open source. [Article 2 of the AI Act](https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-2)\n\n## When Must a Chatbot Disclose That It Is AI?\n\nAI that directly engages in conversation must provide disclosure no later than the first interaction. This may include not only chatbots but also conversational AI agents and avatars. The disclosure must be clearly distinguishable on screen or through audio.\n\nThe Commission’s guidelines assess the following conditions together:\n\n- It qualifies as an AI system under the AI Act.\n- It is designed to engage in substantive two-way exchanges with people.\n- The AI communicates directly, without a person acting as an intermediary.\n- The other party to the conversation is a natural person.\n\nMachine-to-machine communication that no person sees is distinct from this disclosure obligation. Analytical functions operating in the background also do not constitute direct conversation. Whether the obligation applies is not determined merely by whether the system is a simple automated response tool.\n\nA separate disclosure may not be required if it is obvious that the system is AI. The benchmark is a reasonably well-informed, observant, and circumspect ordinary user. The Commission explains that this exception should be interpreted narrowly. [Official Article 50 FAQ](https://digital-strategy.ec.europa.eu/en/faqs/transparency-obligations-under-article-50-ai-act)\n\n## The Difference Between Machine-Readable Marking and On-Screen Labels\n\nMachine-readable marking helps software detect whether content was generated or manipulated by AI. A human-readable on-screen label directly informs users of that fact. The two methods have different purposes and responsible parties.\n\n| Comparison | Machine-readable marking | Human-perceivable disclosure |\n|---|---|---|\n| Main provision | Article 50(2) | Article 50(4) |\n| Primary responsible party | Provider of a generative AI system | Deployer using the relevant content |\n| Scope | Synthetic audio, images, video, and text | Deepfakes and certain public-interest text |\n| Core requirement | Marking and detection must be technically possible | The information must be perceivable without a separate tool |\n| Substitutability | Does not automatically replace human-facing disclosure | Does not automatically replace technical marking |\n\nThe provider’s technology must be robust and reliable to the extent technically feasible. Interoperability with other systems must also be considered. Limitations specific to the type of content and implementation costs are considered as well.\n\nMerely saving something as a JSON file does not satisfy this obligation. A machine’s ability to read a file is different from its ability to detect AI provenance. This distinction follows from the purpose of marking under Article 50. [Article 50 of the AI Act](https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-50)\n\n## Deepfake Labelling Criteria\n\nA deepfake is AI-generated or manipulated content made to appear authentic. The relevant formats are images, audio, and video. It may involve not only people but also objects, places, entities, or events.\n\nThe key considerations are resemblance to the original subject and the possibility of misleading someone about authenticity. Accordingly, not every AI-generated image is automatically a deepfake. Conversely, content is not excluded merely because it does not involve face swapping. The definition is explained in the [Commission’s transparency summary](https://digital-strategy.ec.europa.eu/en/factpages/quick-facts-transparency-rules-ai-systems).\n\nFor deepfakes subject to disclosure, people must be able to recognise the use of AI when they first encounter the content. The method of disclosure may be adjusted for artistic, satirical, or fictional works. In such cases, disclosure may be provided in a way that does not hinder the enjoyment of the work. The disclosure obligation does not disappear entirely merely because the content is an artistic work.\n\nThe labelling icons provided by the EU are optional. Use of an icon does not, by itself, guarantee legal compliance. The disclosure must also satisfy accessibility requirements. [EU labelling icon guidance](https://digital-strategy.ec.europa.eu/en/policies/eu-icons-labelling-ai-generated-content)\n\n\u003e The information shall conform to the applicable accessibility requirements.\n\u003e\n\u003e — EU AI Act Article 50(5)\n\nThis wording expressly states the accessibility requirements applicable to the provision of information. Whether a small icon alone conveys the meaning must be assessed separately. For audio content, consideration must also be given to how listeners perceive the disclosure.\n\n## Is Public-Interest Text Exempt If Reviewed by a Person?\n\nAn exception to disclosure may apply where there is substantive human review and editorial responsibility. It applies to text published to inform the public about matters of public interest. The deployer’s disclosure obligation does not apply to every AI-written document.\n\nThe exception requires the following elements together:\n\n- Human review or substantive editorial control is carried out.\n- A natural or legal person holds editorial responsibility for the publication.\n\nA spell check alone does not constitute substantive review. Human review is a process that assesses the validity of the content. Editorial control involves the authority to approve, revise, or reject the content.\n\nThis exception concerns disclosure of public-interest text. It does not also exempt the provider from the machine-readable marking obligation. Nor does the same exception automatically apply to a deepfake reviewed by a person. [Official FAQ explanation of editorial review](https://digital-strategy.ec.europa.eu/en/faqs/transparency-obligations-under-article-50-ai-act)\n\n## Must Emotion Recognition and Biometric Categorisation Also Be Disclosed?\n\nDeployers must inform affected persons that the relevant system is operating. This obligation is separate from chatbot disclosures or the marking of generated content. Applicable data protection rules also apply to the processing of personal data.\n\nMerely providing disclosure does not make use of the system permissible. Other prohibitions or requirements for high-risk AI may apply separately. Article 50 does not eliminate other regulatory obligations. [Commission guidance on scope](https://digital-strategy.ec.europa.eu/en/policies/guidelines-ai-transparency-obligations)\n\n## Summary by Condition\n\nExceptions must be assessed only within the scope of the relevant obligation. Even for the same content, the conclusions for providers and deployers may differ. The following table compares the official rules by situation.\n\n| Condition or example | Key point in determining applicability |\n|---|---|\n| AI assisting with standard editing | Consider the exception to the machine-readable marking obligation under paragraph 2 |\n| Processing that does not substantially alter input data or its semantics | Consider the exception conditions under paragraph 2 |\n| Substantive review of public-interest text with editorial responsibility | An exception to the deployer’s text disclosure obligation may apply |\n| AI-generated public-interest text published without review | Consider the deployer’s disclosure obligation |\n| Deepfake included in an artistic or satirical work | Disclosure may be provided in a way that does not hinder enjoyment of the work |\n| Use for lawfully authorised crime-prevention or law-enforcement purposes | Verify the purpose and safeguard requirements for each provision |\n\nMerely claiming that a system is used for crime-prevention or law-enforcement purposes does not establish an exception. The use must be legally authorised and satisfy the conditions of the relevant provision. Public-facing AI through which citizens report crimes requires particular caution. Paragraph 1 excludes this case from the law-enforcement exception. [Exceptions by obligation under Article 50](https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-50)\n\n## Common Mistakes About the Application Date and Transition Period\n\nThe transition period until December 2026 is not a transition period for all of Article 50. It is limited to specified systems and obligations. The question of retroactive labelling for existing content must also be considered separately.\n\n| Category | Application date or treatment |\n|---|---|\n| General application of Article 50 | August 2, 2026 |\n| Paragraph 2 obligations for systems placed on the market before August 2, 2026 | Compliance required from December 2, 2026 |\n| Chatbot disclosures and deployers’ deepfake disclosures | The paragraph 2 transition period above does not automatically apply |\n| Content generated before August 2, 2026 | The Commission FAQ explains that there is no retroactive labelling obligation |\n\nThe date a system was placed on the market and the date content was generated are different assessment criteria. New content is not entirely excluded merely because it was produced by an existing system. The schedule is based on the [official FAQ explanation of application dates and transition periods](https://digital-strategy.ec.europa.eu/en/faqs/transparency-obligations-under-article-50-ai-act).\n\n## Labels Disappearing During HTML, JSON, and RSS Republishing\n\nWhen publishing in multiple formats, the label should be checked across each exposure path. Whether a label on a webpage remains in a downloaded file is a separate issue. This section is an operational analysis combining disclosure obligations with guidance on resharing.\n\n| Publication channel | What to check |\n|---|---|\n| HTML page | Is the disclosure perceptible when the content is first viewed? |\n| TXT or Markdown | Does the necessary disclosure remain when separated from the page? |\n| RSS/XML | Is the disclosure conveyed when the feed content is read independently? |\n| JSON or JSON-LD | Is the AI-related information conveyed to users during redisplay? |\n| Image or video download | Does the label remain after downloading and resharing? |\n\nThis does not mean that the law specifies a particular JSON key name. Nor can it be assumed that a data structure alone satisfies the disclosure obligation. The key issue is how content subject to disclosure is presented to people.\n\nThe EU icon guidance addresses visibility during resharing and downloading. Verification by publication channel is therefore a useful operational check. This is distinct from an independent obligation to attach the same label to every data file. [EU icon placement guidance](https://digital-strategy.ec.europa.eu/en/policies/eu-icons-labelling-ai-generated-content)\n\n## The Difference Between Guidelines and a Code of Practice\n\nGuidelines explain the scope of application, while a code of practice helps with implementation methods. The legal basis for the obligations is the AI Act. Participation in a code of practice is voluntary.\n\n| Document | Primary role | Points to note |\n|---|---|---|\n| AI Act | Legal basis for obligations and exceptions | Check application dates and amendments |\n| Article 50 guidelines | Interpretation of concepts, scope, and exceptions | Review against the factual circumstances of the service |\n| Code of Practice on Transparency of AI-Generated Content | Practical implementation of marking, detection, and disclosure | After signing, the relevant measures must actually be implemented |\n\nBusinesses that do not sign the code must still comply with their legal obligations. They may demonstrate that measures implemented through other methods are appropriate. This code of practice is distinct from the code of practice for GPAI models. The former mainly addresses the marking of outputs. [Code of Practice on Transparency of AI-Generated Content](https://digital-strategy.ec.europa.eu/en/policies/code-practice-ai-generated-content)\n\n## Enforcement Authorities and Official Reporting and Complaint Channels\n\nThe reporting channel depends on the authority supervising the system at issue. National competent authorities are responsible for general AI systems. The AI Office is responsible for systems within its statutory jurisdiction. The EDPS is involved with systems used by EU institutions.\n\n| Situation | Official channel to check |\n|---|---|\n| System supervised by a Member State authority | The relevant national market surveillance authority |\n| Suspected violation involving a system under AI Office jurisdiction | AI Act Complaint Tool |\n| Information about a violation learned through a work-related relationship | AI Act Whistleblower Tool |\n| Related complaint by a provider integrating an external GPAI model | GPAI complaint channel for downstream providers |\n\nReporting channels can be distinguished in the following order:\n\n1. Determine whether the subject at issue is an AI system or a GPAI model.\n2. Check jurisdiction in the Commission’s guidance on the scope of supervision.\n3. Proceed to the complaint or reporting channel designated by the relevant authority.\n\nRelevant corporate violations may result in substantial financial penalties. The Commission states that the maximum may be up to €15 million. Another benchmark is 3% of worldwide annual turnover in the preceding financial year. For ordinary companies, the higher of the two amounts is the maximum.\n\nThe actual penalty is determined according to factors including the nature, gravity, and duration of the infringement. Separate proportionality considerations apply to SMEs and others. Article 50 violations are not subject to a uniform fixed penalty. [Enforcement framework and official reporting channels](https://digital-strategy.ec.europa.eu/en/policies/enforcement-ai-act)\n\n## Frequently Asked Questions\n\n### Does All AI-Generated Content Need a Visible Label?\n\nThe deployer’s disclosure obligation does not apply uniformly to all generated content. Deepfakes and certain public-interest text are the main categories covered. The provider’s machine-readable marking obligation must be assessed separately.\n\n### Must a Chatbot Repeat the Disclosure Every Time It Responds?\n\nUnder Article 50, disclosure must be provided no later than the first interaction. The rule does not require the same wording to be repeated in every sentence. However, a first-time user must be able to perceive it clearly.\n\n### Does an AI-Written Article Reviewed by a Person Need No Label?\n\nThe assessment does not end with substantive review or editorial control alone. There must also be an entity that holds editorial responsibility for the publication. It is difficult to apply this exception based on a spell check alone.\n\n### Is a Satirical Video Exempt from Deepfake Labelling?\n\nSatire does not provide a complete exemption by itself. If the content qualifies as a deepfake, the method of disclosure may be adjusted. A method that does not hinder the display or enjoyment of the work is permitted.\n\n### Does Using an EU Icon Guarantee Compliance?\n\nUse of an icon alone does not guarantee legal compliance. Its use is optional. The timing, perceptibility, and accessibility of the label must also be assessed.\n\n### Are All Obligations Deferred Until December 2026?\n\nThe transition period is limited to the Article 50(2) obligations for existing systems. Those systems must comply from December 2, 2026. It is not a blanket transition period for chatbot disclosures and deepfake disclosures.\n\n### Does an AI Label Mean the Content Is Factual?\n\nAn AI label provides information about whether content was generated or manipulated. It does not certify fact-checking or copyright permission. The accuracy and legality of the content must be assessed separately.","content_html":"\u003cp\u003eArticle 50 of the EU AI Act applies from August 2, 2026. Disclosures for chatbots and deepfakes are central. Machine-readable marking of generated content is also required. However, the responsible party and the exceptions and transition periods differ by obligation.\u003c/p\u003e\n\u003cp\u003eThe dates and figures in this article are based on the Commission’s official guidance from July and August 2026.\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#what-does-article-50-require\" class=\"anchor\" id=\"what-does-article-50-require\"\u003e\u003c/a\u003eWhat Does Article 50 Require?\u003c/h2\u003e\n\u003cp\u003eArticle 50 sets out transparency rules requiring disclosure of AI use. Not every obligation applies to every AI system. The applicable provision depends on the system’s function and the purpose of the content.\u003c/p\u003e\n\u003cdiv class=\"overflow-x-auto\"\u003e\u003ctable\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth\u003eApplicable situation\u003c/th\u003e\n\u003cth\u003ePrimary responsible party\u003c/th\u003e\n\u003cth\u003eRequired action\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Applicable situation\"\u003eAI that interacts directly with people\u003c/td\u003e\n\u003ctd data-label=\"Primary responsible party\"\u003eProvider\u003c/td\u003e\n\u003ctd data-label=\"Required action\"\u003eDesign it to disclose that the person is interacting with AI\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Applicable situation\"\u003eAI that generates audio, images, video, or text\u003c/td\u003e\n\u003ctd data-label=\"Primary responsible party\"\u003eProvider\u003c/td\u003e\n\u003ctd data-label=\"Required action\"\u003eEnsure machine-readable marking and detectability of outputs\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Applicable situation\"\u003eEmotion recognition or biometric categorisation system\u003c/td\u003e\n\u003ctd data-label=\"Primary responsible party\"\u003eDeployer\u003c/td\u003e\n\u003ctd data-label=\"Required action\"\u003eInform affected persons that the system is operating\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Applicable situation\"\u003eDeepfake or public-interest text subject to disclosure\u003c/td\u003e\n\u003ctd data-label=\"Primary responsible party\"\u003eDeployer\u003c/td\u003e\n\u003ctd data-label=\"Required action\"\u003eDisclose that the content was generated or manipulated by AI\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\u003c/div\u003e\n\u003cp\u003eThe Commission published interpretive guidelines on July 20, 2026. In an announcement on July 31 of the same year, it reaffirmed that the rules would apply from August 2. This date does not apply uniformly to every provision of the AI Act. \u003ca href=\"https://digital-strategy.ec.europa.eu/en/news/commission-starts-enforcing-ai-act-rules-and-new-transparency-requirements-2-august\"\u003eCommission enforcement announcement\u003c/a\u003e\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#comparing-the-responsibilities-of-providers-and-deployers\" class=\"anchor\" id=\"comparing-the-responsibilities-of-providers-and-deployers\"\u003e\u003c/a\u003eComparing the Responsibilities of Providers and Deployers\u003c/h2\u003e\n\u003cp\u003eResponsibility is divided according to the role actually performed, rather than the company’s industry. A provider is an entity that develops an AI system or has one developed. It places the system on the market or puts it into service under its own name or trademark. A deployer is an entity that uses an AI system under its authority.\u003c/p\u003e\n\u003cdiv class=\"overflow-x-auto\"\u003e\u003ctable\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth\u003eCategory\u003c/th\u003e\n\u003cth\u003eProvider\u003c/th\u003e\n\u003cth\u003eDeployer\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Category\"\u003eStatutory term\u003c/td\u003e\n\u003ctd data-label=\"Provider\"\u003eprovider\u003c/td\u003e\n\u003ctd data-label=\"Deployer\"\u003edeployer\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Category\"\u003eTypical role\u003c/td\u003e\n\u003ctd data-label=\"Provider\"\u003eProvides an AI system under its own name\u003c/td\u003e\n\u003ctd data-label=\"Deployer\"\u003eUses AI to produce content or perform work\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Category\"\u003eCore obligation\u003c/td\u003e\n\u003ctd data-label=\"Provider\"\u003eDesign that enables disclosure and technical marking\u003c/td\u003e\n\u003ctd data-label=\"Deployer\"\u003eDisclosure appropriate to the context of use and exposure\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Category\"\u003eBasis for determination\u003c/td\u003e\n\u003ctd data-label=\"Provider\"\u003eResponsibility for development, market placement, or putting into service\u003c/td\u003e\n\u003ctd data-label=\"Deployer\"\u003eAuthority over the use of the system\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\u003c/div\u003e\n\u003cp\u003eIn this article, “deployer” means deployer. It is a different concept from a distributor involved in sales and distribution. A single organisation may act as both provider and deployer. The definitions are based on \u003ca href=\"https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-3\"\u003eArticle 3 of the AI Act\u003c/a\u003e.\u003c/p\u003e\n\u003cp\u003eBusinesses outside the EU may also fall within its scope. This includes cases where an AI system is provided in the EU market. It also applies where the system’s output is used in the EU. Being based in South Korea does not, by itself, exclude a business.\u003c/p\u003e\n\u003cp\u003eDeployer obligations do not apply to purely personal, non-professional use. Business or professional activities are distinct from this exception. Article 50 is not exempted merely because a system is free or open source. \u003ca href=\"https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-2\"\u003eArticle 2 of the AI Act\u003c/a\u003e\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#when-must-a-chatbot-disclose-that-it-is-ai\" class=\"anchor\" id=\"when-must-a-chatbot-disclose-that-it-is-ai\"\u003e\u003c/a\u003eWhen Must a Chatbot Disclose That It Is AI?\u003c/h2\u003e\n\u003cp\u003eAI that directly engages in conversation must provide disclosure no later than the first interaction. This may include not only chatbots but also conversational AI agents and avatars. The disclosure must be clearly distinguishable on screen or through audio.\u003c/p\u003e\n\u003cp\u003eThe Commission’s guidelines assess the following conditions together:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eIt qualifies as an AI system under the AI Act.\u003c/li\u003e\n\u003cli\u003eIt is designed to engage in substantive two-way exchanges with people.\u003c/li\u003e\n\u003cli\u003eThe AI communicates directly, without a person acting as an intermediary.\u003c/li\u003e\n\u003cli\u003eThe other party to the conversation is a natural person.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eMachine-to-machine communication that no person sees is distinct from this disclosure obligation. Analytical functions operating in the background also do not constitute direct conversation. Whether the obligation applies is not determined merely by whether the system is a simple automated response tool.\u003c/p\u003e\n\u003cp\u003eA separate disclosure may not be required if it is obvious that the system is AI. The benchmark is a reasonably well-informed, observant, and circumspect ordinary user. The Commission explains that this exception should be interpreted narrowly. \u003ca href=\"https://digital-strategy.ec.europa.eu/en/faqs/transparency-obligations-under-article-50-ai-act\"\u003eOfficial Article 50 FAQ\u003c/a\u003e\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#the-difference-between-machine-readable-marking-and-on-screen-labels\" class=\"anchor\" id=\"the-difference-between-machine-readable-marking-and-on-screen-labels\"\u003e\u003c/a\u003eThe Difference Between Machine-Readable Marking and On-Screen Labels\u003c/h2\u003e\n\u003cp\u003eMachine-readable marking helps software detect whether content was generated or manipulated by AI. A human-readable on-screen label directly informs users of that fact. The two methods have different purposes and responsible parties.\u003c/p\u003e\n\u003cdiv class=\"overflow-x-auto\"\u003e\u003ctable\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth\u003eComparison\u003c/th\u003e\n\u003cth\u003eMachine-readable marking\u003c/th\u003e\n\u003cth\u003eHuman-perceivable disclosure\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Comparison\"\u003eMain provision\u003c/td\u003e\n\u003ctd data-label=\"Machine-readable marking\"\u003eArticle 50(2)\u003c/td\u003e\n\u003ctd data-label=\"Human-perceivable disclosure\"\u003eArticle 50(4)\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Comparison\"\u003ePrimary responsible party\u003c/td\u003e\n\u003ctd data-label=\"Machine-readable marking\"\u003eProvider of a generative AI system\u003c/td\u003e\n\u003ctd data-label=\"Human-perceivable disclosure\"\u003eDeployer using the relevant content\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Comparison\"\u003eScope\u003c/td\u003e\n\u003ctd data-label=\"Machine-readable marking\"\u003eSynthetic audio, images, video, and text\u003c/td\u003e\n\u003ctd data-label=\"Human-perceivable disclosure\"\u003eDeepfakes and certain public-interest text\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Comparison\"\u003eCore requirement\u003c/td\u003e\n\u003ctd data-label=\"Machine-readable marking\"\u003eMarking and detection must be technically possible\u003c/td\u003e\n\u003ctd data-label=\"Human-perceivable disclosure\"\u003eThe information must be perceivable without a separate tool\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Comparison\"\u003eSubstitutability\u003c/td\u003e\n\u003ctd data-label=\"Machine-readable marking\"\u003eDoes not automatically replace human-facing disclosure\u003c/td\u003e\n\u003ctd data-label=\"Human-perceivable disclosure\"\u003eDoes not automatically replace technical marking\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\u003c/div\u003e\n\u003cp\u003eThe provider’s technology must be robust and reliable to the extent technically feasible. Interoperability with other systems must also be considered. Limitations specific to the type of content and implementation costs are considered as well.\u003c/p\u003e\n\u003cp\u003eMerely saving something as a JSON file does not satisfy this obligation. A machine’s ability to read a file is different from its ability to detect AI provenance. This distinction follows from the purpose of marking under Article 50. \u003ca href=\"https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-50\"\u003eArticle 50 of the AI Act\u003c/a\u003e\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#deepfake-labelling-criteria\" class=\"anchor\" id=\"deepfake-labelling-criteria\"\u003e\u003c/a\u003eDeepfake Labelling Criteria\u003c/h2\u003e\n\u003cp\u003eA deepfake is AI-generated or manipulated content made to appear authentic. The relevant formats are images, audio, and video. It may involve not only people but also objects, places, entities, or events.\u003c/p\u003e\n\u003cp\u003eThe key considerations are resemblance to the original subject and the possibility of misleading someone about authenticity. Accordingly, not every AI-generated image is automatically a deepfake. Conversely, content is not excluded merely because it does not involve face swapping. The definition is explained in the \u003ca href=\"https://digital-strategy.ec.europa.eu/en/factpages/quick-facts-transparency-rules-ai-systems\"\u003eCommission’s transparency summary\u003c/a\u003e.\u003c/p\u003e\n\u003cp\u003eFor deepfakes subject to disclosure, people must be able to recognise the use of AI when they first encounter the content. The method of disclosure may be adjusted for artistic, satirical, or fictional works. In such cases, disclosure may be provided in a way that does not hinder the enjoyment of the work. The disclosure obligation does not disappear entirely merely because the content is an artistic work.\u003c/p\u003e\n\u003cp\u003eThe labelling icons provided by the EU are optional. Use of an icon does not, by itself, guarantee legal compliance. The disclosure must also satisfy accessibility requirements. \u003ca href=\"https://digital-strategy.ec.europa.eu/en/policies/eu-icons-labelling-ai-generated-content\"\u003eEU labelling icon guidance\u003c/a\u003e\u003c/p\u003e\n\u003cblockquote\u003e\n\u003cp\u003eThe information shall conform to the applicable accessibility requirements.\u003c/p\u003e\n\u003cp\u003e— EU AI Act Article 50(5)\u003c/p\u003e\n\u003c/blockquote\u003e\n\u003cp\u003eThis wording expressly states the accessibility requirements applicable to the provision of information. Whether a small icon alone conveys the meaning must be assessed separately. For audio content, consideration must also be given to how listeners perceive the disclosure.\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#is-public-interest-text-exempt-if-reviewed-by-a-person\" class=\"anchor\" id=\"is-public-interest-text-exempt-if-reviewed-by-a-person\"\u003e\u003c/a\u003eIs Public-Interest Text Exempt If Reviewed by a Person?\u003c/h2\u003e\n\u003cp\u003eAn exception to disclosure may apply where there is substantive human review and editorial responsibility. It applies to text published to inform the public about matters of public interest. The deployer’s disclosure obligation does not apply to every AI-written document.\u003c/p\u003e\n\u003cp\u003eThe exception requires the following elements together:\u003c/p\u003e\n\u003cul\u003e\n\u003cli\u003eHuman review or substantive editorial control is carried out.\u003c/li\u003e\n\u003cli\u003eA natural or legal person holds editorial responsibility for the publication.\u003c/li\u003e\n\u003c/ul\u003e\n\u003cp\u003eA spell check alone does not constitute substantive review. Human review is a process that assesses the validity of the content. Editorial control involves the authority to approve, revise, or reject the content.\u003c/p\u003e\n\u003cp\u003eThis exception concerns disclosure of public-interest text. It does not also exempt the provider from the machine-readable marking obligation. Nor does the same exception automatically apply to a deepfake reviewed by a person. \u003ca href=\"https://digital-strategy.ec.europa.eu/en/faqs/transparency-obligations-under-article-50-ai-act\"\u003eOfficial FAQ explanation of editorial review\u003c/a\u003e\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#must-emotion-recognition-and-biometric-categorisation-also-be-disclosed\" class=\"anchor\" id=\"must-emotion-recognition-and-biometric-categorisation-also-be-disclosed\"\u003e\u003c/a\u003eMust Emotion Recognition and Biometric Categorisation Also Be Disclosed?\u003c/h2\u003e\n\u003cp\u003eDeployers must inform affected persons that the relevant system is operating. This obligation is separate from chatbot disclosures or the marking of generated content. Applicable data protection rules also apply to the processing of personal data.\u003c/p\u003e\n\u003cp\u003eMerely providing disclosure does not make use of the system permissible. Other prohibitions or requirements for high-risk AI may apply separately. Article 50 does not eliminate other regulatory obligations. \u003ca href=\"https://digital-strategy.ec.europa.eu/en/policies/guidelines-ai-transparency-obligations\"\u003eCommission guidance on scope\u003c/a\u003e\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#summary-by-condition\" class=\"anchor\" id=\"summary-by-condition\"\u003e\u003c/a\u003eSummary by Condition\u003c/h2\u003e\n\u003cp\u003eExceptions must be assessed only within the scope of the relevant obligation. Even for the same content, the conclusions for providers and deployers may differ. The following table compares the official rules by situation.\u003c/p\u003e\n\u003cdiv class=\"overflow-x-auto\"\u003e\u003ctable\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth\u003eCondition or example\u003c/th\u003e\n\u003cth\u003eKey point in determining applicability\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Condition or example\"\u003eAI assisting with standard editing\u003c/td\u003e\n\u003ctd data-label=\"Key point in determining applicability\"\u003eConsider the exception to the machine-readable marking obligation under paragraph 2\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Condition or example\"\u003eProcessing that does not substantially alter input data or its semantics\u003c/td\u003e\n\u003ctd data-label=\"Key point in determining applicability\"\u003eConsider the exception conditions under paragraph 2\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Condition or example\"\u003eSubstantive review of public-interest text with editorial responsibility\u003c/td\u003e\n\u003ctd data-label=\"Key point in determining applicability\"\u003eAn exception to the deployer’s text disclosure obligation may apply\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Condition or example\"\u003eAI-generated public-interest text published without review\u003c/td\u003e\n\u003ctd data-label=\"Key point in determining applicability\"\u003eConsider the deployer’s disclosure obligation\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Condition or example\"\u003eDeepfake included in an artistic or satirical work\u003c/td\u003e\n\u003ctd data-label=\"Key point in determining applicability\"\u003eDisclosure may be provided in a way that does not hinder enjoyment of the work\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Condition or example\"\u003eUse for lawfully authorised crime-prevention or law-enforcement purposes\u003c/td\u003e\n\u003ctd data-label=\"Key point in determining applicability\"\u003eVerify the purpose and safeguard requirements for each provision\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\u003c/div\u003e\n\u003cp\u003eMerely claiming that a system is used for crime-prevention or law-enforcement purposes does not establish an exception. The use must be legally authorised and satisfy the conditions of the relevant provision. Public-facing AI through which citizens report crimes requires particular caution. Paragraph 1 excludes this case from the law-enforcement exception. \u003ca href=\"https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-50\"\u003eExceptions by obligation under Article 50\u003c/a\u003e\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#common-mistakes-about-the-application-date-and-transition-period\" class=\"anchor\" id=\"common-mistakes-about-the-application-date-and-transition-period\"\u003e\u003c/a\u003eCommon Mistakes About the Application Date and Transition Period\u003c/h2\u003e\n\u003cp\u003eThe transition period until December 2026 is not a transition period for all of Article 50. It is limited to specified systems and obligations. The question of retroactive labelling for existing content must also be considered separately.\u003c/p\u003e\n\u003cdiv class=\"overflow-x-auto\"\u003e\u003ctable\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth\u003eCategory\u003c/th\u003e\n\u003cth\u003eApplication date or treatment\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Category\"\u003eGeneral application of Article 50\u003c/td\u003e\n\u003ctd data-label=\"Application date or treatment\"\u003eAugust 2, 2026\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Category\"\u003eParagraph 2 obligations for systems placed on the market before August 2, 2026\u003c/td\u003e\n\u003ctd data-label=\"Application date or treatment\"\u003eCompliance required from December 2, 2026\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Category\"\u003eChatbot disclosures and deployers’ deepfake disclosures\u003c/td\u003e\n\u003ctd data-label=\"Application date or treatment\"\u003eThe paragraph 2 transition period above does not automatically apply\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Category\"\u003eContent generated before August 2, 2026\u003c/td\u003e\n\u003ctd data-label=\"Application date or treatment\"\u003eThe Commission FAQ explains that there is no retroactive labelling obligation\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\u003c/div\u003e\n\u003cp\u003eThe date a system was placed on the market and the date content was generated are different assessment criteria. New content is not entirely excluded merely because it was produced by an existing system. The schedule is based on the \u003ca href=\"https://digital-strategy.ec.europa.eu/en/faqs/transparency-obligations-under-article-50-ai-act\"\u003eofficial FAQ explanation of application dates and transition periods\u003c/a\u003e.\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#labels-disappearing-during-html-json-and-rss-republishing\" class=\"anchor\" id=\"labels-disappearing-during-html-json-and-rss-republishing\"\u003e\u003c/a\u003eLabels Disappearing During HTML, JSON, and RSS Republishing\u003c/h2\u003e\n\u003cp\u003eWhen publishing in multiple formats, the label should be checked across each exposure path. Whether a label on a webpage remains in a downloaded file is a separate issue. This section is an operational analysis combining disclosure obligations with guidance on resharing.\u003c/p\u003e\n\u003cdiv class=\"overflow-x-auto\"\u003e\u003ctable\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth\u003ePublication channel\u003c/th\u003e\n\u003cth\u003eWhat to check\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Publication channel\"\u003eHTML page\u003c/td\u003e\n\u003ctd data-label=\"What to check\"\u003eIs the disclosure perceptible when the content is first viewed?\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Publication channel\"\u003eTXT or Markdown\u003c/td\u003e\n\u003ctd data-label=\"What to check\"\u003eDoes the necessary disclosure remain when separated from the page?\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Publication channel\"\u003eRSS/XML\u003c/td\u003e\n\u003ctd data-label=\"What to check\"\u003eIs the disclosure conveyed when the feed content is read independently?\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Publication channel\"\u003eJSON or JSON-LD\u003c/td\u003e\n\u003ctd data-label=\"What to check\"\u003eIs the AI-related information conveyed to users during redisplay?\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Publication channel\"\u003eImage or video download\u003c/td\u003e\n\u003ctd data-label=\"What to check\"\u003eDoes the label remain after downloading and resharing?\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\u003c/div\u003e\n\u003cp\u003eThis does not mean that the law specifies a particular JSON key name. Nor can it be assumed that a data structure alone satisfies the disclosure obligation. The key issue is how content subject to disclosure is presented to people.\u003c/p\u003e\n\u003cp\u003eThe EU icon guidance addresses visibility during resharing and downloading. Verification by publication channel is therefore a useful operational check. This is distinct from an independent obligation to attach the same label to every data file. \u003ca href=\"https://digital-strategy.ec.europa.eu/en/policies/eu-icons-labelling-ai-generated-content\"\u003eEU icon placement guidance\u003c/a\u003e\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#the-difference-between-guidelines-and-a-code-of-practice\" class=\"anchor\" id=\"the-difference-between-guidelines-and-a-code-of-practice\"\u003e\u003c/a\u003eThe Difference Between Guidelines and a Code of Practice\u003c/h2\u003e\n\u003cp\u003eGuidelines explain the scope of application, while a code of practice helps with implementation methods. The legal basis for the obligations is the AI Act. Participation in a code of practice is voluntary.\u003c/p\u003e\n\u003cdiv class=\"overflow-x-auto\"\u003e\u003ctable\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth\u003eDocument\u003c/th\u003e\n\u003cth\u003ePrimary role\u003c/th\u003e\n\u003cth\u003ePoints to note\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Document\"\u003eAI Act\u003c/td\u003e\n\u003ctd data-label=\"Primary role\"\u003eLegal basis for obligations and exceptions\u003c/td\u003e\n\u003ctd data-label=\"Points to note\"\u003eCheck application dates and amendments\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Document\"\u003eArticle 50 guidelines\u003c/td\u003e\n\u003ctd data-label=\"Primary role\"\u003eInterpretation of concepts, scope, and exceptions\u003c/td\u003e\n\u003ctd data-label=\"Points to note\"\u003eReview against the factual circumstances of the service\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Document\"\u003eCode of Practice on Transparency of AI-Generated Content\u003c/td\u003e\n\u003ctd data-label=\"Primary role\"\u003ePractical implementation of marking, detection, and disclosure\u003c/td\u003e\n\u003ctd data-label=\"Points to note\"\u003eAfter signing, the relevant measures must actually be implemented\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\u003c/div\u003e\n\u003cp\u003eBusinesses that do not sign the code must still comply with their legal obligations. They may demonstrate that measures implemented through other methods are appropriate. This code of practice is distinct from the code of practice for GPAI models. The former mainly addresses the marking of outputs. \u003ca href=\"https://digital-strategy.ec.europa.eu/en/policies/code-practice-ai-generated-content\"\u003eCode of Practice on Transparency of AI-Generated Content\u003c/a\u003e\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#enforcement-authorities-and-official-reporting-and-complaint-channels\" class=\"anchor\" id=\"enforcement-authorities-and-official-reporting-and-complaint-channels\"\u003e\u003c/a\u003eEnforcement Authorities and Official Reporting and Complaint Channels\u003c/h2\u003e\n\u003cp\u003eThe reporting channel depends on the authority supervising the system at issue. National competent authorities are responsible for general AI systems. The AI Office is responsible for systems within its statutory jurisdiction. The EDPS is involved with systems used by EU institutions.\u003c/p\u003e\n\u003cdiv class=\"overflow-x-auto\"\u003e\u003ctable\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth\u003eSituation\u003c/th\u003e\n\u003cth\u003eOfficial channel to check\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Situation\"\u003eSystem supervised by a Member State authority\u003c/td\u003e\n\u003ctd data-label=\"Official channel to check\"\u003eThe relevant national market surveillance authority\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Situation\"\u003eSuspected violation involving a system under AI Office jurisdiction\u003c/td\u003e\n\u003ctd data-label=\"Official channel to check\"\u003eAI Act Complaint Tool\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Situation\"\u003eInformation about a violation learned through a work-related relationship\u003c/td\u003e\n\u003ctd data-label=\"Official channel to check\"\u003eAI Act Whistleblower Tool\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd data-label=\"Situation\"\u003eRelated complaint by a provider integrating an external GPAI model\u003c/td\u003e\n\u003ctd data-label=\"Official channel to check\"\u003eGPAI complaint channel for downstream providers\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\u003c/div\u003e\n\u003cp\u003eReporting channels can be distinguished in the following order:\u003c/p\u003e\n\u003col\u003e\n\u003cli\u003eDetermine whether the subject at issue is an AI system or a GPAI model.\u003c/li\u003e\n\u003cli\u003eCheck jurisdiction in the Commission’s guidance on the scope of supervision.\u003c/li\u003e\n\u003cli\u003eProceed to the complaint or reporting channel designated by the relevant authority.\u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eRelevant corporate violations may result in substantial financial penalties. The Commission states that the maximum may be up to €15 million. Another benchmark is 3% of worldwide annual turnover in the preceding financial year. For ordinary companies, the higher of the two amounts is the maximum.\u003c/p\u003e\n\u003cp\u003eThe actual penalty is determined according to factors including the nature, gravity, and duration of the infringement. Separate proportionality considerations apply to SMEs and others. Article 50 violations are not subject to a uniform fixed penalty. \u003ca href=\"https://digital-strategy.ec.europa.eu/en/policies/enforcement-ai-act\"\u003eEnforcement framework and official reporting channels\u003c/a\u003e\u003c/p\u003e\n\u003ch2\u003e\n\u003ca href=\"#frequently-asked-questions\" class=\"anchor\" id=\"frequently-asked-questions\"\u003e\u003c/a\u003eFrequently Asked Questions\u003c/h2\u003e\n\u003ch3\u003e\n\u003ca href=\"#does-all-ai-generated-content-need-a-visible-label\" class=\"anchor\" id=\"does-all-ai-generated-content-need-a-visible-label\"\u003e\u003c/a\u003eDoes All AI-Generated Content Need a Visible Label?\u003c/h3\u003e\n\u003cp\u003eThe deployer’s disclosure obligation does not apply uniformly to all generated content. Deepfakes and certain public-interest text are the main categories covered. The provider’s machine-readable marking obligation must be assessed separately.\u003c/p\u003e\n\u003ch3\u003e\n\u003ca href=\"#must-a-chatbot-repeat-the-disclosure-every-time-it-responds\" class=\"anchor\" id=\"must-a-chatbot-repeat-the-disclosure-every-time-it-responds\"\u003e\u003c/a\u003eMust a Chatbot Repeat the Disclosure Every Time It Responds?\u003c/h3\u003e\n\u003cp\u003eUnder Article 50, disclosure must be provided no later than the first interaction. The rule does not require the same wording to be repeated in every sentence. However, a first-time user must be able to perceive it clearly.\u003c/p\u003e\n\u003ch3\u003e\n\u003ca href=\"#does-an-ai-written-article-reviewed-by-a-person-need-no-label\" class=\"anchor\" id=\"does-an-ai-written-article-reviewed-by-a-person-need-no-label\"\u003e\u003c/a\u003eDoes an AI-Written Article Reviewed by a Person Need No Label?\u003c/h3\u003e\n\u003cp\u003eThe assessment does not end with substantive review or editorial control alone. There must also be an entity that holds editorial responsibility for the publication. It is difficult to apply this exception based on a spell check alone.\u003c/p\u003e\n\u003ch3\u003e\n\u003ca href=\"#is-a-satirical-video-exempt-from-deepfake-labelling\" class=\"anchor\" id=\"is-a-satirical-video-exempt-from-deepfake-labelling\"\u003e\u003c/a\u003eIs a Satirical Video Exempt from Deepfake Labelling?\u003c/h3\u003e\n\u003cp\u003eSatire does not provide a complete exemption by itself. If the content qualifies as a deepfake, the method of disclosure may be adjusted. A method that does not hinder the display or enjoyment of the work is permitted.\u003c/p\u003e\n\u003ch3\u003e\n\u003ca href=\"#does-using-an-eu-icon-guarantee-compliance\" class=\"anchor\" id=\"does-using-an-eu-icon-guarantee-compliance\"\u003e\u003c/a\u003eDoes Using an EU Icon Guarantee Compliance?\u003c/h3\u003e\n\u003cp\u003eUse of an icon alone does not guarantee legal compliance. Its use is optional. The timing, perceptibility, and accessibility of the label must also be assessed.\u003c/p\u003e\n\u003ch3\u003e\n\u003ca href=\"#are-all-obligations-deferred-until-december-2026\" class=\"anchor\" id=\"are-all-obligations-deferred-until-december-2026\"\u003e\u003c/a\u003eAre All Obligations Deferred Until December 2026?\u003c/h3\u003e\n\u003cp\u003eThe transition period is limited to the Article 50(2) obligations for existing systems. Those systems must comply from December 2, 2026. It is not a blanket transition period for chatbot disclosures and deepfake disclosures.\u003c/p\u003e\n\u003ch3\u003e\n\u003ca href=\"#does-an-ai-label-mean-the-content-is-factual\" class=\"anchor\" id=\"does-an-ai-label-mean-the-content-is-factual\"\u003e\u003c/a\u003eDoes an AI Label Mean the Content Is Factual?\u003c/h3\u003e\n\u003cp\u003eAn AI label provides information about whether content was generated or manipulated. It does not certify fact-checking or copyright permission. The accuracy and legality of the content must be assessed separately.\u003c/p\u003e\n","tags":["EU AI Act","Generative AI","Personal data protection","AI Development","Deepfake"],"faqs":[{"question":"When does Article 50 of the EU AI Act become applicable?","answer":"In principle, it applies from August 2, 2026. The labeling and detection obligations under paragraph 2 are an exception for systems released earlier. Those obligations must be complied with from December 2, 2026."},{"question":"Can Korean companies also be subject to Article 50?","answer":"They may be subject to it if they provide AI systems in the EU market. It is also relevant if the systems' outputs are used in the EU. Being located in Korea alone does not exclude them from its application."},{"question":"Are machine-readable markings and human-readable labels the same thing?","answer":"They are methods for fulfilling different obligations. Machine-readable markings facilitate the technical detection of whether content was AI-generated or manipulated. Disclosure to people enables users to recognize that fact."},{"question":"Are all AI images deepfakes?","answer":"Not all AI images are deepfakes. Resemblance to a real subject and the potential to mislead people about authenticity are factors in the assessment. Content depicting objects, places, or events, as well as people, may also be relevant."},{"question":"Does merely having a person check the spelling in an AI-generated article qualify for a disclosure exemption?","answer":"A spelling check alone does not constitute substantive review. Human review of the content or editorial control is required. There must also be an entity that assumes editorial responsibility for the publication."},{"question":"Does including JSON-LD satisfy the machine-readable marking obligation?","answer":"The use of JSON-LD alone does not determine whether the obligation is satisfied. It must be possible to indicate and detect whether content was AI-generated or manipulated. The data format and the legal marking requirements must be distinguished."},{"question":"Must deepfakes created for artistic or satirical purposes also be disclosed?","answer":"They are not fully exempt merely because they are intended for artistic or satirical purposes. Adjustments to the method of disclosure are permitted for such works. The disclosure may be made in an appropriate manner that does not interfere with the enjoyment of the work."},{"question":"Is the EU disclosure icon mandatory?","answer":"Use of the EU icon is optional. The disclosure obligation itself is not optional. Merely displaying the icon does not guarantee compliance."},{"question":"Is joining the Code of Practice on transparency of generated content mandatory?","answer":"Joining the Code of Practice is voluntary. Businesses that do not join must still comply with the obligations under the AI Act. They may demonstrate compliance by other appropriate means."},{"question":"Should all suspected violations of Article 50 be reported to the AI Office?","answer":"The reporting channel varies depending on the system's supervisory authority. The AI Office complaint tool is the channel for systems under its jurisdiction. For other systems, the competent national authority of the Member State must be identified."}],"sources":[{"url":"https://digital-strategy.ec.europa.eu/en/library/guidelines-transparency-obligations-providers-and-deployers-ai-systems","title":"European Commission — Guidelines on transparency obligations for providers and deployers of AI systems","type":"source"},{"url":"https://digital-strategy.ec.europa.eu/en/news/commission-starts-enforcing-ai-act-rules-and-new-transparency-requirements-2-august","title":"European Commission — Commission starts enforcing AI Act rules and new transparency requirements on 2 August","type":"source"},{"url":"https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-50","title":"AI Act Service Desk — Article 50: Transparency obligations","type":"source"},{"url":"https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-3","title":"AI Act Service Desk — Article 3: Definitions","type":"source"},{"url":"https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-2","title":"AI Act Service Desk — Article 2: Scope","type":"source"},{"url":"https://digital-strategy.ec.europa.eu/en/faqs/transparency-obligations-under-article-50-ai-act","title":"European Commission — Transparency obligations under Article 50 of the AI Act","type":"source"},{"url":"https://digital-strategy.ec.europa.eu/en/factpages/quick-facts-transparency-rules-ai-systems","title":"European Commission — Quick Facts: Transparency rules for AI systems","type":"source"},{"url":"https://digital-strategy.ec.europa.eu/en/policies/guidelines-ai-transparency-obligations","title":"European Commission — Guidelines on transparency obligations for providers and deployers of certain AI systems","type":"source"},{"url":"https://digital-strategy.ec.europa.eu/en/policies/code-practice-ai-generated-content","title":"European Commission — Code of Practice on Transparency of AI-generated Content","type":"source"},{"url":"https://digital-strategy.ec.europa.eu/en/policies/eu-icons-labelling-ai-generated-content","title":"European Commission — EU Icons for labelling AI-generated content","type":"source"},{"url":"https://digital-strategy.ec.europa.eu/en/policies/enforcement-ai-act","title":"European Commission — The enforcement framework of the AI Act","type":"source"}],"images":[{"id":1352,"url":"https://injoys.com/rails/active_storage/blobs/proxy/eyJfcmFpbHMiOnsiZGF0YSI6MTk0MzAsInB1ciI6ImJsb2JfaWQifX0=--5be6c9d0722472b53fc1d9645580bcafc01b6a8b/ai-5b5ba769.webp","is_representative":true,"generation_method":"ai_photo","license":"ai_generated","mime_type":"image/webp","width":1536,"height":1024,"translations":{"ko":{"alt":"사무실에서 여성 진행자 영상을 편집하는 작업자와 영상 편집 화면","caption":"작업자가 모니터로 여성 진행자 영상을 검토하며 편집하고 있다.","description":null},"en":{"alt":"Editor reviewing footage of a female presenter on a video editing monitor","caption":"An editor reviews and edits presenter footage at her workstation.","description":null},"ja":{"alt":"オフィスで女性司会者の映像を編集する作業者と編集画面","caption":"作業者がモニターで女性司会者の映像を確認しながら編集している。","description":null},"es":{"alt":"Editora revisando el video de una presentadora en un monitor de edición","caption":"Una editora revisa y monta el video de una presentadora en su puesto de trabajo.","description":null},"id":{"alt":"Editor meninjau video pembawa acara perempuan pada monitor penyuntingan","caption":"Seorang editor meninjau dan menyunting video pembawa acara di meja kerjanya.","description":null},"pt":{"alt":"Editora revisando o vídeo de uma apresentadora em um monitor de edição","caption":"Uma editora revisa e edita o vídeo de uma apresentadora em sua estação de trabalho.","description":null},"zh-hant":{"alt":"工作人員在辦公室的剪輯螢幕上檢視女性主持人影片","caption":"工作人員在工作站上檢視並剪輯女性主持人的影片。","description":null},"de":{"alt":"Editorin prüft das Video einer Moderatorin auf einem Schnittmonitor","caption":"Eine Editorin sichtet und bearbeitet das Video einer Moderatorin an ihrem Arbeitsplatz.","description":null}}},{"id":1353,"url":"https://injoys.com/rails/active_storage/blobs/proxy/eyJfcmFpbHMiOnsiZGF0YSI6MTk0MzYsInB1ciI6ImJsb2JfaWQifX0=--01eafc8a334015e605eb18531fc73bf3c3ad4b74/ai-88eea99c.webp","is_representative":false,"generation_method":"ai_semi","license":"ai_generated","mime_type":"image/webp","width":1536,"height":1024,"translations":{"ko":{"alt":"영사실에서 AI 아이콘이 표시된 영상 편집 화면을 조작하는 여성","caption":"한 영상 편집자가 AI 생성 콘텐츠의 표시 요소를 화면에서 확인하고 있다.","description":null},"en":{"alt":"Woman editing video on a monitor displaying AI and chatbot icons in a projection room","caption":"A video editor reviews on-screen labels for AI-generated content.","description":null},"ja":{"alt":"映写室でAIとチャットボットのアイコンが映る動画編集画面を操作する女性","caption":"映像編集者がAI生成コンテンツの表示要素を画面で確認している。","description":null},"es":{"alt":"Mujer editando un video con iconos de IA y chatbot en una sala de proyección","caption":"Una editora revisa en pantalla indicadores de contenido generado por IA.","description":null},"id":{"alt":"Perempuan mengedit video dengan ikon AI dan chatbot di ruang proyeksi","caption":"Seorang editor video meninjau penanda konten buatan AI pada layar.","description":null},"pt":{"alt":"Mulher editando vídeo com ícones de IA e chatbot em uma sala de projeção","caption":"Uma editora revisa na tela indicadores de conteúdo gerado por IA.","description":null},"zh-hant":{"alt":"女子在放映室操作顯示人工智慧與聊天機器人圖示的影片剪輯畫面","caption":"影片剪輯師正在螢幕上檢查人工智慧生成內容的標示元素。","description":null},"de":{"alt":"Frau bearbeitet in einem Vorführraum ein Video mit KI- und Chatbot-Symbolen","caption":"Eine Videoeditorin prüft Kennzeichnungen für KI-generierte Inhalte auf dem Bildschirm.","description":null}}}],"published_at":"2026-09-17T17:06:07+09:00","updated_at":"2026-09-17T17:06:07+09:00","license":"cc_by","translation_status":"reviewed","available_locales":["ko","en","ja","es"],"data_locales":["ko","en","ja","es","id","pt","zh-hant","de"],"url":"https://injoys.com/en/articles/eu-ai-act-article-50-chatbot-deepfake-transparency"}