On 20 July 2026, the EU has introduced a simple visual language for identifying certain AI-generated and AI-modified content and with the intent to simplify compliance with the transparency requirements under the EU AI Act[1].
Moreover, starting 2 August 2026, Article 50 of the EU AI Act enters into force and requires that providers build transparency into certain AI systems, whilst are now required to disclose deepfakes and certain AI-generated public-interest text. For AI systems that were already on the market before 2 August 2026, a grace period has been decided regarding the obligations on the part of AI providers to have the content generated through their AI systems marked in a machine-readable format, with compliance to this specific requirement being shifted to 2 December 2026.
The three icons suggested by the EU to facilitate the labelling of deepfakes
The Commission proposes three choices of icons. Each suggested icon is intended to help properly disclose a specific use of AI.
These icons are not intended to help compliance with every transparency requirement under the AI Act. These icons only help deployers fulfil their obligation to label the deepfakes and the public-interest text that falls within Article 50(4). Using an icon does not, by itself, establish compliance with every transparency requirement.
 |
 |
 |
| BASIC ICON
To be used when AI was involved in the creation of deep fake content (image, audio, video) or published text or when a custom text label or interactive second layer is implemented |
FULLY AI-GENERATED
To be used when the entire deep fake content (image, audio, video) or text is fully generated by AI with no human-created elements or human editorial control (apart from prompting) |
PARTIALLY AI-MODIFIED
To be used when pre-existing, human-made content was partially modified with AI turning it into deepfake content or a text on matters of public interest |
Examples of use include:
- Authentic photographs of an empty apartment are manipulated using AI = “AI MODIFIED”; and
- AI-generated news summaries = “AI GENERATED”.
How do transparency requirements under the AI Act apply to AI providers and AI deployers?
Article 50 distinguishes technical marking and general disclosure by the providers of AI systems on one hand, and disclosure to the person seeing or hearing the content generated by the deployers of AI systems on the other hand.
| PROVIDERS: BUILD TRANSPARENCY IN THE AI SYSTEMS | DEPLOYERS: DISCLOSE USE OF AI TO PEOPLE |
| People must be informed when they interact directly with AI, unless that would be obvious to a user who was reasonably well-informed, observant and circumspect, taking into account the circumstances and the context of use. | The appropriate information of natural persons and the processing of personal data in accordance with Regulation (EU) 2016/679[2] (GDPR) and Regulation (EU) 2018/1725[3] (protection of data by EU institutions), as well as Directive (EU) 2016/680[4] (protection of data by competent authorities), whenever an emotion recognition system or a biometric categorisation system is deployed |
| Synthetic audio, image, video and text outputs must carry effective, robust and interoperable machine-readable marks, subject to limited exceptions. | Disclosure to AI users whenever content (image, audio or video) constitutes deep fake as well as when a text has been generated or manipulated via AI. Disclosure by the deployers can be done via the icons described hereabove. |
Important limits and exceptions to transparency obligation
Human-reviewed text: Public-interest text does not require labelling by deployers where a knowledgeable person deliberately reviews its substance or an editorial entity can approve, alter or reject it, and a natural or legal person assumes editorial responsibility. Spell-checking or a purely formal review is not enough.
Creative works: For evidently artistic, creative, satirical, fictional or analogous deepfake content, disclosure is still required, but the transparency requirements will be attenuated so as not to hamper the display or enjoyment of the deepfake content.
Personal use: Purely personal, non-professional use of AI by a person falls outside the deployer definition and is thus not subject to transparency requirements.
Standard editing: The obligation on the part of a provider to provide AI systems marked in a machine-readable format does not apply where the AI systems merely perform a standard editing function or do not substantially alter the input or its meaning.
Law enforcement: Narrow exceptions also apply to uses authorised by law for detecting, preventing, investigating or prosecuting criminal offences.
What to do now
Deployers of AI systems are advised to make use of the icons suggested by the EU in order to simplify their disclosure obligations when deploying deepfake content and AI-generated or manipulated texts.
Moreover, since the transparency requirements under the AI Act are now applicable, providers and deployers of AI systems should ensure that they are compliant with said transparency requirements and that they have taken into account the guidance provided in the Guidelines published by the European Commission on 20 July 2026[5].
We must stress that, whilst compliance with the transparency requirements of the AI Act may seem complicated to achieve in practice, voluntary adherence to the Code of Practice on Transparency of AI-Generated Content which was published in June 2026 will also help demonstrate compliance to users of AI system and competent authorities alike.
[1] Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence
[2] Regulation (EU) 2016/67 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data
[3] Regulation (EU) 2018/1725 on the protection of natural persons with regard to the processing of personal data by the Union institutions, bodies, offices and agencies and on the free movement of such data
[4] Directive (EU) 2016/680 on the protection of natural persons with regard to the processing of personal data by competent authorities for the purposes of the prevention, investigation, detection or prosecution of criminal offences or the execution of criminal penalties, and on the free movement of such data
[5] Guidelines on the implementation of the transparency obligations for certain AI systems underArticle 50 of Regulation (EU) 2024/1689 (the ‘AI Act’)
