What was passed
In spring 2026 the Ministry of Digital Development proposed a tough version: mandatory labelling of any image, video and audio generated by AI. In the end the text was softened.
On 8 July 2026 the State Duma passed the law on the development of artificial intelligence technologies, and on 26 July the president signed it. The law establishes the basic concepts in the field of AI and the conditions for developing domestic technology, while the question of labelling was settled gently.
The main point: labelling is voluntary
For an ordinary user and for a business, the key provision reads like this: people who create audio and video with AI may, at their own discretion, place a warning that the content is generated.
So you have no obligation to label every generated picture or clip.
What platforms must do
The obligation in the law falls on large services. Owners of sites, social networks, information systems and apps with a daily audience of more than 500 thousand users must give users a technical way to put such a label on their content.
In practice this means that large social networks and video platforms will eventually have a button or tag «created with AI». Using it is not mandatory under the law, but it will be there.
When it takes effect
The provisions on labelling AI content take effect on 1 March 2027. Until then platforms have time to build the necessary features.
| What | When |
|---|---|
| Law passed by the State Duma | 8 July 2026 |
| Signed by the president | 26 July 2026 |
| Labelling provisions take effect | 1 March 2027 |
Do not confuse it with labelling advertising
This is the most important practical clarification, and it is easy to get wrong.
The voluntary rule concerns the «created by AI» label. Labelling online advertising is a separate obligation under the advertising law, and it has not gone anywhere. If you use a generated clip or banner as an ad, it must be labelled as an ad, like any other advertising creative. How the picture was made does not affect this obligation.
Roughly:
- You generated a picture for a post on your own blog: an «AI» label is optional.
- You generated a clip and are running it as an ad: ad labelling is mandatory, an «AI» label is optional.
Platform rules are a separate layer
Apart from the law, many platforms have their own rules on synthetic content. Some large video platforms, for example, ask you to disclose realistic generated or altered content, especially if it could be mistaken for real footage. The rules change, so check them separately for each platform where you publish.
Is it worth labelling if it is not mandatory
My practical opinion: often it is.
The reason is not legal but reputational. If a viewer could take a generated person, event or «review» for real and later learns the truth, they stop trusting the rest of your content. An honest label costs less than restoring trust.
Where a label is especially appropriate:
- realistic people whom a viewer could take for real ones;
- clips that look like documentary footage;
- product images where the surroundings, not the product itself, were generated.
Where it matters less: obvious graphics, illustrations, cartoons, abstraction, backgrounds.
And separately: with real people in the frame, it is not only labelling that matters but also the protection of a citizen’s image. More in the article on deepfakes and the law.
What a business should do now
- Label ads as ads, regardless of what the creative was made with.
- Check the rules of the platforms where you publish: they may have their own requirements.
- Decide on an internal policy: where you label AI content voluntarily. One rule for the whole team is better than decisions on the fly.
- Keep your generation history. It will be useful both in disputes and in questions of authorship; more in the article on copyright for AI images.
- Follow the changes. AI regulation changes fast, and clarifications may appear by March 2027.
What else to check before launching an AI ad: AI video for commercial purposes.
In short
The law on AI development was signed on 26 July 2026. Labelling generated content is voluntary for users, and large platforms with a daily audience of 500 thousand or more must provide a technical way to add such a label. The provisions take effect on 1 March 2027. At the same time, labelling advertising remains mandatory regardless of how the creative was made, and platforms may have their own rules. This is not legal advice: in disputed cases it is better to check with a lawyer.
