--- title: "EU AI Act Article 50: AI Video Transparency and VideoElephant Submission Requirements for Content Providers" date: 2026-07-30T10:45:00-07:00 author: "Denise O'Keeffe" canonical_url: "https://videoelephant.com/blog/eu-ai-act-ai-generated-video-requirements" section: Blog --- ###### In this post... [EU AI Act Article 50: Key facts](#eu-ai-act-article-50-key-facts) [What is Article 50 of the EU AI Act?](#what-is-article-50-of-the-eu-ai-act) [When do the EU AI content-transparency rules take effect?](#when-do-the-eu-ai-content-transparency-rules-take-effect) [Which AI-generated videos may need to be disclosed?](#which-ai-generated-videos-may-need-to-be-disclosed) [What is considered a deepfake under the EU AI Act?](#what-is-considered-a-deepfake-under-the-eu-ai-act) [Do all videos created with AI need to be labelled?](#do-all-videos-created-with-ai-need-to-be-labelled) [How does the standard-editing exception apply?](#how-does-the-standard-editing-exception-apply) [How should VideoElephant content providers identify affected content?](#how-should-video-elephant-content-providers-identify-affected-content) [What are Content Credentials and C2PA metadata?](#what-are-content-credentials-and-c2-pa-metadata) [How should artistic, fictional or satirical AI content be disclosed?](#how-should-artistic-fictional-or-satirical-ai-content-be-disclosed) [Does the EU AI Act require older videos to be labelled retroactively?](#does-the-eu-ai-act-require-older-videos-to-be-labelled-retroactively) [What is the EU Code of Practice on Transparency of AI-Generated Content?](#what-is-the-eu-code-of-practice-on-transparency-of-ai-generated-content) [Why is VideoElephant requesting AI information from content providers?](#why-is-video-elephant-requesting-ai-information-from-content-providers) [What should a content provider do when uncertain?](#what-should-a-content-provider-do-when-uncertain) [Frequently Asked Questions](#frequently-asked-questions) [Official EU resources](#official-eu-resources) [Our commitment to transparent video distribution](#our-commitment-to-transparent-video-distribution) Back to top For video content, the rules are primarily intended to help audiences identify realistic synthetic or manipulated material that could otherwise be mistaken for authentic footage. VideoElephant requires its content providers to identify material uses of generative AI in newly submitted content. This supports our own ingestion and distribution processes and helps us respond to publisher and platform requirements. #### EU AI Act Article 50: Key facts - The EU AI Act entered into force on August 1, 2024. - The Article 50 transparency requirements apply from August 2, 2026. - Providers of generative AI systems have responsibilities relating to the machine-readable marking and detection of AI-generated or manipulated content. - Professional users of AI systems, described as deployers under the Act, must disclose deepfakes and certain other covered content. - Routine AI-assisted editing does not automatically make a video a deepfake or require a viewer-facing label. - The European Commission’s current guidance indicates that AI-generated or manipulated content generated before August 2, 2026 does not require mandatory retroactive labelling, although voluntary labelling is encouraged - The EU Code of Practice on Transparency of AI-Generated Content is voluntary, but the underlying Article 50 obligations are legal requirements. #### What is Article 50 of the EU AI Act? Article 50 establishes transparency obligations for certain types of AI systems and AI-generated content. The provisions most relevant to video creation and distribution are Article 50(2) and Article 50(4). Article 50(2) applies to providers of AI systems that generate synthetic audio, image, video or text. These providers must ensure that covered outputs are marked in a machine-readable format and can be detected as artificially generated or manipulated, as far as technically feasible. Article 50(4) applies to deployers using AI systems for professional purposes. It requires them to disclose when image, audio or video content constitutes a deepfake. The European Commission’s current guidance indicates that AI-generated or manipulated content generated before August 2, 2026 does not require mandatory retroactive labelling, although voluntary labelling is encouraged #### When do the EU AI content-transparency rules take effect? The relevant Article 50 transparency obligations apply from August 2, 2026. The EU AI Act itself entered into force on August 1, 2024, but its requirements have been introduced according to a phased implementation timetable. Content providers should review their workflows before August 2, 2026 and ensure that appropriate information, disclosures and provenance data can be supplied with newly created content. #### Which AI-generated videos may need to be disclosed? AI-generated or AI-manipulated video may require disclosure when it realistically depicts a person, place, object, organization or event and could falsely appear authentic or truthful. The EU AI Act refers to this type of content as a deepfake. Examples may include: - An AI-generated presenter or realistic digital person - A cloned or synthetic voice presented as a real person’s voice - A face swap - AI-generated lip-sync manipulation - A real person appearing to say or do something that did not occur - Synthetic footage depicting a real location, organization or event - AI-generated scenes presented as authentic news or documentary footage - Material alterations that could cause viewers to misunderstand what happened Whether a specific video falls within Article 50 depends on its content, context, presentation and the role of the organization creating or using it. #### What is considered a deepfake under the EU AI Act? A deepfake is AI-generated or AI-manipulated image, audio or video content that resembles existing people, objects, places, entities or events and could falsely appear authentic or truthful. The term is not limited to deceptive impersonations of celebrities or public figures. It can also include realistic synthetic representations of organizations, locations, products or events. A clearly animated character or an obviously fictional visual effect may be less likely to be mistaken for authentic footage. A realistic AI presenter, reconstructed event or cloned voice requires closer assessment. #### Do all videos created with AI need to be labelled? No. The EU AI Act does not create a blanket requirement to label every video in which an AI tool was used. The focus of the relevant video-disclosure obligation is realistic AI-generated or manipulated content that could falsely appear authentic. Article 50(2) contains an exception relating to AI systems used for standard editing or which do not substantially alter the input data or its meaning. Whether that exception applies depends on the particular technology, content and use. However, VideoElephant may request information about AI use beyond the minimum legal disclosure required for a particular video. This allows us to evaluate the content against contractual, publisher, platform and territory-specific requirements. #### How does the standard-editing exception apply? Whether the standard-editing exception applies depends on how the AI was used and whether it substantially altered the source content or its meaning. Examples that may fall within the exception include basic colour correction, noise reduction, stabilisation, caption transcription and faithful text translation, where they do not materially generate or alter the content. These examples are illustrative only. Content providers remain responsible for making their own assessment and should obtain independent legal advice where necessary #### How should VideoElephant content providers identify affected content? As part of VideoElephant’s submission requirements, content providers must use the applicable exact lowercase tags: Please provide: - Where all or a material part of the video or audio was created using generative AI, including synthetic presenters or voices. - Manipulated: where existing video or audio was materially altered using AI, including face swaps, lip-sync manipulation or generative reconstruction. - Ai-disclosure-included: where a visible or audible AI disclosure is already included in the content. - Where more than one description applies, include all relevant tags. If neither ai-generated nor ai-manipulated is included, VideoElephant will treat the submission as confirmation that the content contains no material generative AI use requiring disclosure to us. VideoElephant operational tags and are not prescribed by the EU AI Act. They do not replace any viewer-facing disclosure or technical marking that may be required. Content providers should also preserve provenance metadata and Content Credentials contained in the original file and should not intentionally remove them before delivery. #### What are Content Credentials and C2PA metadata? Content Credentials provide information about the origin and editing history of digital media. They may identify how a piece of content was created, which tools were used and whether AI-generated or manipulated elements are present. The Coalition for Content Provenance and Authenticity, or C2PA, maintains a technical standard used to attach cryptographically verifiable provenance information to digital content. The EU AI Act does not require every content provider to use one named commercial standard. However, Article 50 includes machine-readable marking and detection requirements for providers of covered generative AI systems. VideoElephant asks providers to preserve Content Credentials, C2PA data and other provenance information where they are available. #### How should artistic, fictional or satirical AI content be disclosed? AI-generated material within an evidently artistic, creative, satirical, fictional or similar work may still require disclosure. Article 50 permits the disclosure to be made in an appropriate manner that does not interfere with the display or enjoyment of the work. Depending on the content, examples may include: - A brief notice at the beginning of the video; - A clear on-screen label or icon when the relevant content first appears; or - An audible disclosure before or when the relevant content is first presented. These examples are illustrative only. A disclosure included solely in end credits, embedded metadata or a description may not be sufficient because it may not be clear to the audience at the time of first exposure. Content providers remain responsible for determining the appropriate disclosure for their content and should obtain independent legal advice where necessary. #### Does the EU AI Act require older videos to be labelled retroactively? The European Commission’s current guidance indicates that AI-generated or manipulated content generated before 2 August 2026 does not require mandatory retroactive labelling, although voluntary labelling is encouraged. Examples include: - Generated and published before 2 August 2026: current guidance indicates that no retroactive AI label is required. - Generated before 2 August 2026 but submitted to VideoElephant after that date: a retroactive viewer-facing label may not be required, but VideoElephant’s applicable AI tags must still be included with the submission. - Generated on or after 2 August 2026: the content should be assessed under the applicable Article 50 requirements. - Older content materially regenerated or manipulated using AI after 2 August 2026: the new or altered elements should be assessed under the applicable Article 50 requirements. These examples are illustrative only. Content providers remain responsible for confirming how the applicable transitional rules apply to their content and should obtain independent legal advice where necessary. #### What is the EU Code of Practice on Transparency of AI-Generated Content? The Code of Practice on Transparency of AI-Generated Content is a voluntary framework designed to help providers and deployers comply with Article 50. The European Commission published the Code on June 10, 2026. It contains two principal sections: 1. Guidance for providers on marking and detecting AI-generated or manipulated content 2. Guidance for deployers on labelling deepfakes and certain AI-generated or manipulated text On July 8, 2026, the European Commission concluded that the Code adequately covers the relevant obligations under Article 50. The European AI Board adopted its adequacy assessment the following day. Organizations are not legally required to sign the Code. However, signatories may use its measures as a recognized framework for demonstrating compliance. Organizations that do not sign remain responsible for demonstrating compliance through other appropriate measures. Adherence to the Code does not replace the EU AI Act or guarantee compliance in every individual case. #### Why is VideoElephant requesting AI information from content providers? VideoElephant distributes video content to publishers, platforms and audiences across multiple territories. Receiving accurate information about generative AI use helps us: - Assess whether content may require disclosure or labelling - Preserve audience trust - Respond to publisher and platform requirements - Maintain accurate content metadata - Protect provenance information - Support transparent global distribution - Evaluate whether content is suitable for a particular territory, platform or context VideoElephant’s information request may be broader than the minimum Article 50 requirement for a specific video. It is intended to create a consistent submission process across our global partner network. #### What should a content provider do when uncertain? When there is uncertainty, content providers should disclose the AI use to VideoElephant rather than assume that it is outside the requirements. Please contact the VideoElephant Content Team before submitting content when: - A realistic person, voice, location or event was generated - AI materially reconstructed missing visual or audio material - Synthetic dubbing or voice cloning was used - The content blends authentic and synthetic footage - A viewer could reasonably believe that synthetic content is real - The appropriate label or metadata is unclear VideoElephant can review the information for its own ingestion and distribution purposes but cannot determine a content provider’s legal compliance or provide legal advice on its behalf. Each content provider remains responsible for obtaining its own legal advice and ensuring compliance with applicable laws, rights, contracts and platform requirements. #### Frequently Asked Questions **What date does Article 50 of the EU AI Act apply?** The relevant transparency obligations under Article 50 apply from August 2, 2026. **Does every AI-generated video need a visible label?** No. Article 50(4) requires deployers to disclose image, audio or video content that constitutes a deepfake. Separate machine-readable marking obligations apply to providers of covered generative AI systems. Whether particular content falls within these provisions requires a fact-specific assessment. **Do AI-generated presenters need to be disclosed?** A realistic AI-generated presenter should generally be disclosed to VideoElephant. A viewer-facing disclosure may also be required when the presenter could be mistaken for a real person or authentic footage. **Does an AI-cloned voice need to be disclosed?** Yes. Synthetic or cloned voices should be disclosed to VideoElephant, particularly when the voice represents or resembles a real person. **Do AI captions need to be disclosed?** For VideoElephant’s tagging purposes, automated transcription alone does not generally require an AI tag where it accurately represents the source and does not materially generate or alter the content. This is an operational submission rule, not a determination of legal compliance. **Does AI translation need to be disclosed?** For VideoElephant’s tagging purposes, faithful text translation alone does not generally require an AI tag. Synthetic dubbing, voice cloning or translation that materially changes the speaker’s words, identity or meaning should be tagged. This is an operational submission rule, not a determination of legal compliance. **Does AI upscaling need to be disclosed?** For VideoElephant’s tagging purposes, technical upscaling alone does not generally require an AI tag. Generative reconstruction that creates realistic new details not present in the source material should be tagged. This is an operational submission rule, not a determination of legal compliance. **Do older AI-generated videos need retroactive labels?** The European Commission’s current guidance indicates that content generated before August 2, 2026 does not require mandatory retroactive labelling, although voluntary labelling is encouraged. Content providers should confirm how the applicable transitional position applies to their content. **Is the EU Code of Practice mandatory?** No. Signing the Code is voluntary. Compliance with the applicable Article 50 obligations is mandatory, and organizations that do not follow the Code must demonstrate compliance through other adequate measures. **Is C2PA required by the EU AI Act?** The Act does not require every content provider to use C2PA specifically. It does require providers of covered generative AI systems to support machine-readable marking and detection. C2PA and Content Credentials are examples of provenance technologies that may support transparency. **Who is responsible for compliance?** Responsibility under Article 50 depends principally on the role performed. An organization that provides a generative AI system may be a provider, while an organization that uses an AI system professionally may be a deployer. Merely creating, owning rights in or distributing completed content does not necessarily determine an organization’s status. Content providers remain responsible for their own legal and contractual compliance. #### Official EU resources For additional information, content providers should consult the following official European Union resources: - [European Commission: EU AI Act overview](https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai) - [European Commission: Code of Practice on Transparency of AI-Generated Content](https://digital-strategy.ec.europa.eu/en/policies/code-practice-ai-generated-content) - [European Commission FAQ: Code of Practice on Transparency of AI-Generated Content](https://digital-strategy.ec.europa.eu/en/faqs/code-practice-transparency-ai-generated-content) - [European Commission FAQ: Signing the Code and retroactive labelling](https://digital-strategy.ec.europa.eu/en/faqs/signing-code-practice-transparency-ai-generated-content) - [European Commission: Assessment of the Code of Practice](https://digital-strategy.ec.europa.eu/en/library/commission-opinion-assessment-code-practice-transparency-ai-generated-content) - [Official text of Regulation (EU) 2024/1689 on EUR-Lex](https://eur-lex.europa.eu/eli/reg/2024/1689/oj/eng) #### Our commitment to transparent video distribution VideoElephant works with content creators, rights holders, publishers and platforms around the world. Clear information about generative AI will help us maintain audience trust, preserve content provenance and respond appropriately to evolving regulatory and platform requirements. Content providers who are uncertain whether a particular AI use should be disclosed should contact the VideoElephant Content Team before submitting the content. *This post is provided for general informational purposes only and does not constitute legal advice. VideoElephant’s tagging and submission requirements are operational requirements for content supplied to us; they do not determine whether particular content complies with the EU AI Act. Content providers remain responsible for assessing their own legal obligations and should obtain independent legal advice where necessary.* ![Kate Doughan](https://assets.videoelephant.com/team-members/_156x156_crop_center-center_none/16999/doughan.webp)###### Kate Doughan Head of Content ###### Kate Doughan Head of Content Kate sources premium video globally for VideoElephant. Transitioning from diverse sales roles to acquisitions in 2011, she holds a BA in English and Art History and a Masters in Media and International Conflict (2010). [ ](https://www.linkedin.com/in/kate-doughan-1a979334/) ###### In this post... [EU AI Act Article 50: Key facts](#eu-ai-act-article-50-key-facts) [What is Article 50 of the EU AI Act?](#what-is-article-50-of-the-eu-ai-act) [When do the EU AI content-transparency rules take effect?](#when-do-the-eu-ai-content-transparency-rules-take-effect) [Which AI-generated videos may need to be disclosed?](#which-ai-generated-videos-may-need-to-be-disclosed) [What is considered a deepfake under the EU AI Act?](#what-is-considered-a-deepfake-under-the-eu-ai-act) [Do all videos created with AI need to be labelled?](#do-all-videos-created-with-ai-need-to-be-labelled) [How does the standard-editing exception apply?](#how-does-the-standard-editing-exception-apply) [How should VideoElephant content providers identify affected content?](#how-should-video-elephant-content-providers-identify-affected-content) [What are Content Credentials and C2PA metadata?](#what-are-content-credentials-and-c2-pa-metadata) [How should artistic, fictional or satirical AI content be disclosed?](#how-should-artistic-fictional-or-satirical-ai-content-be-disclosed) [Does the EU AI Act require older videos to be labelled retroactively?](#does-the-eu-ai-act-require-older-videos-to-be-labelled-retroactively) [What is the EU Code of Practice on Transparency of AI-Generated Content?](#what-is-the-eu-code-of-practice-on-transparency-of-ai-generated-content) [Why is VideoElephant requesting AI information from content providers?](#why-is-video-elephant-requesting-ai-information-from-content-providers) [What should a content provider do when uncertain?](#what-should-a-content-provider-do-when-uncertain) [Frequently Asked Questions](#frequently-asked-questions) [Official EU resources](#official-eu-resources) [Our commitment to transparent video distribution](#our-commitment-to-transparent-video-distribution) Back to top ###### Stay up to date Subscribe for our latest updates and video insights #### You may also like [ ![How VideoElephant Is Scaling AI, FAST Channels, and CTV Out of Home](https://assets.videoelephant.com/images/_435x238_crop_center-center_none/46368/Brian-beet-tv-blog.webp) ###### How VideoElephant Is Scaling AI, FAST Channels, and CTV Out of Home At CES earlier this year, VideoElephant’s Chief Commercial Officer sat down with Beet.TV to discuss how video distribution, AI, FAST channels, and CTV Out of Home are reshaping the media ecosystem. ![Brian Cullinane](https://assets.videoelephant.com/team-members/_200x200_crop_center-center_none/4953/bio-1.webp) Brian Cullinane Feb 13, 2026 ](https://videoelephant.com/blog/how-videoelephant-is-scaling-ai-fast-channels-and-ctv-out-of-home) [ ![Publishers Accelerate Video Growth Amid AI Search Disruption: Vertical Video and Creator Syndication Drive Record Revenue Gains, New VideoElephant Data Shows](https://assets.videoelephant.com/images/_435x238_crop_center-center_none/43882/Rewind-cover-press-release-2.webp) ###### Publishers Accelerate Video Growth Amid AI Search Disruption: Vertical Video and Creator Syndication Drive Record Revenue Gains, New VideoElephant Data Shows New York, NY, November 11, 2025 – VideoElephant’s October 2025 Rewind Report reveals a transformative surge in licensed and monetized video content usage among top global publishers, as AI-driven search reshapes digital media distribution. ![Victoria Vitarelli](https://assets.videoelephant.com/team-members/_200x200_crop_center-center_none/16970/vitarelli.webp) Victoria Vitarelli Nov 11, 2025 ](https://videoelephant.com/blog/videoelephant-rewind-report-2025)