The VIFLY guide France · European Union
A clear AI disclosure, suited to your content.
A proofreader, an invented realistic scene and a news article raise different questions. Start with what changed and what your audience might believe.
Sources checked on 23 October 2026
The short answer
Required disclosures
The AI Act does not require a visible label on every piece of AI-assisted content. It requires disclosure for deepfakes and certain text published on matters of public interest. French advertising rules and platform policies may add requirements.
These guideposts distinguish the status of a rule; 'Pause and check' calls for review before acting.
Three questions, in this order
What does the audience see?
An authentic scene or a creation that could be mistaken for one?
Which rule applies?
AI Act, commercial influence, contract and platform.
Where should you explain it?
Understandable information when the content is first encountered.
From rule to practice
What does that look like in practice?
A city panorama showing an invented flood
Viewers could mistake it for real news even with no recognizable person.
Disclose generation, state that it is an illustration and check platform rules.
A document proofread and reviewed by its author
The tool fixes spelling without changing meaning or inventing a scene.
Do not infer a blanket labeling duty. Check contractual and editorial context and remain responsible for the content.
An AI-written news article published without review
The text informs the public on a matter of public interest with no editorial review.
Disclosure is required in this context. Also establish genuine verification before publishing.
Machine-readable marking and visible disclosure
The provider of a generative system must enable machine-readable marking of covered outputs. A professional using the system must inform exposed people in the specified cases. Metadata or a watermark therefore does not automatically replace an explanation for the audience.
Article 50 obligations have applied since 2 August 2026. The transition until 2 December 2026 concerns only provider marking and detection for systems placed on the market before 2 August; it is not a general delay to professional users’ disclosures.
Images, voices and video: the risk of false authenticity
A deepfake can resemble an existing person, object, place or event and falsely appear authentic. Do not judge an invented realistic scene only by whether a famous face appears. Disclose its artificial nature clearly and accessibly, no later than first exposure.
For an evidently artistic, satirical or fictional work, disclosure can be adapted to preserve presentation. This does not remove every information duty, personal right or rule of the platform distributing the work.
Public-interest text: the editorial exception
AI-generated or manipulated text published to inform the public on a matter of public interest has a specific disclosure rule. The exception requires human review or editorial control and a natural or legal person taking editorial responsibility. Adding a byline alone does not document that work.
Proofreading or drafting assistance does not automatically trigger every rule. Keep sources, verify facts and explain your process when the context requires it. Information still needs fact-checking, with or without AI.
Choose a disclosure people understand
Put the information where people encounter the content: on the image, beside the player, in the text or in the relevant disclosure setting. Avoid unreadable labels, buried hashtags or explanations several clicks away. Check accessibility needs and the medium’s rules.
EU icons are an available tool, not a universally mandatory logo for every creation. The code of practice is voluntary; the applicable legal duty is not. An icon should clarify AI’s role without suggesting legal certification.
Wording to get started
Disclosures to adapt to your case.
These examples are neither official forms nor a compliance guarantee. Check the wording, placement and display duration required. French statutory disclosures are retained in the examples.
Illustrative scene
AI-generated scene for illustration. It does not depict an actual event.
Synthetic voice
AI-generated voice. This person did not speak this recording.
Editorial transparency
This content was prepared with AI assistance, then checked and edited under our editorial team’s responsibility.
Text copied. Adapt it before use.Copy unavailable: select the text to copy it.
Before taking action
Your essential checks.
A preparation aid, with no automatic compliance verdict.
Getting to the point
Questions people often ask.
Must every image say it is AI-generated?
There is no single blanket duty covering every image. Check deepfake, advertising, rights and platform context: several rules can apply together.
Is a watermark enough to inform the audience?
Not automatically. A technical signal and clear information for people are different layers.
Does 2 December 2026 apply to every creator?
No. This transition concerns Article 50(2) and certain already marketed systems, not a general extension for creator disclosures.
Sources and method
The sources behind the answers.
Editorial team: VIFLY. Documentary checks on 23 October 2026. This guide is an educational summary without individual legal validation. The links let you consult the text and its conditions.
- European CommissionAI Act — Article 50: transparency Legal text
- European CommissionGuidelines on transparency obligations Official explanation
- European CommissionQuestions and answers on Article 50 and dates Official explanation
- European CommissionEuropean icons for labelling AI content Official explanation
- European CommissionCode of practice on AI content transparency Official explanation
- European CommissionPractical reference points for AI system transparency Official explanation
- French Ministry of EconomyAI transparency: measures applicable since August 2026 Official explanation
Understand, then practise
Clear rules. Useful applications.
Explore AI courses and VIFLY resources to build a more considered practice.