Blog
Artificial Intelligence is already part of the routine of many marketing and communications teams, and is widely used to create images, adapt photographs, develop videos, test concepts or prepare a first version of a text.
However, since 2 August 2026, there have been new obligations under the AI Act, the European Artificial Intelligence Regulation, which brands must comply with.
In this regard, some questions arise:
What changed on 2 August?
The AI Act entered into force in 2024, but since 2 August 2026, new transparency obligations related to the use of Artificial Intelligence have applied.
In the case of content, these rules cover, among other situations:
For marketing and social media teams, the issue of deepfakes is very important.
But what exactly is a deepfake?
When we talk about deepfakes, we usually think of fake videos of politicians, celebrities or other public figures, but the concept is broader.
It can be an image, video or audio created or altered with Artificial Intelligence that depicts a person, a place, an object or an event so realistically that it can be mistaken for something real.
Does an image created by AI always have to be identified?
In the case of an illustration, with imaginary visual elements, no one expects that image to be a real photograph. But if it is a hyper-realistic image of a person, created entirely with AI and published as if it were a photograph of a client, employee or real situation, the situation changes.
If there is a possibility that someone viewing the post may interpret that image as real, there is a much greater likelihood that we are dealing with content that should be identified.
The context in which the image appears also matters.
An image may be understood differently if it appears in a clearly creative campaign or if it is published as if it documented a real event.
Do we have to identify everything that is made with AI?
The AI Act does not require brands to add the indication “AI-generated/AI-modified” to all content that may have been created with the help of AI.
The obligation depends on the type of content, the way AI was used and, above all, the possibility of that content being interpreted as real or authentic.
“AI-generated” or “AI-modified”?
The AI Act does not require all organisations to use exactly these two expressions.
However, for communication purposes, there are two simple ways to make the use of AI understandable to the public.
A brand may, for example, internally adopt the following distinction:
AI-generated
This can be used when the image or video has been created entirely through Artificial Intelligence.
AI-modified
This can be used when we start with real content, but AI has introduced significant changes.
The European Union has also made icons available to support a more harmonious identification of content created or altered through AI. But it is not mandatory for all brands to use exactly these expressions and icons. The most important thing is that, whenever identification is necessary, it is clear and easy to understand.
Where should this identification appear?
An image can be saved, sent by message or shared in another context. Therefore, integrating the indication directly into the image can be a good way to ensure that this information is not lost.
In cases where it is necessary to identify the content, the indication must be visible and perceptible.
It should not be hidden among hashtags, in a comment or in an area that probably no one will see.
In an image, for example, it can be discreetly integrated into one of the corners of the piece.
In a video or Reel, it should appear visibly and for long enough to be read.
And when we use AI only to edit a photograph?
Here it is important to distinguish a minor edit from a significant alteration.
Using an AI tool to improve the lighting, reduce noise, increase resolution, make small corrections or improve image quality, is not the same as, using AI to add people who were never in that location, remove important elements, alter settings, create a situation that never happened or significantly change what the photograph represents.
In the first case, we are much closer to a normal edit., Whereas inbut in the second, we may be changing the very reality represented in the image.
Not everything involving AI needs a label
This is probably the most important point.
AI can be present at various stages of the creative process without this requiring “AI-generated” to be added to every piece.
It can be used to research ideas, test visual directions, improve images, adapt formats or prepare first versions of texts.
Therefore, instead of asking only “did we use AI?”, we need to ask another question:“
Did AI create or alter something that could lead the public to believe they are looking at a real situation?”
If the answer is yes, it is time to determine whether the content falls under the transparency rules and, in that case, identify it.
And what about content published before 2 August?
The new obligations began to apply on 2 August 2026.
This means that brands should take these rules into account for new content published, precisely from that date onwards.
There is also a transitional period until 2 December 2026, but it is important to understand what this means.
This deadline is related to certain technical obligations of AI system providers.
It does not mean that brands can simply wait until December to start complying with the transparency rules that already apply.
In practice, the recommendation is simple: these rules should be considered from now on.
Transparency does not mean stopping the use of AI
The AI Act does not prevent brands from using Artificial Intelligence.
What changes is the need for greater transparency when the technology creates content that can be mistaken for reality.
Ultimately, the simplest rule may be this:
If it looks real, but was created or altered artificially, it is better to stop and determine whether it should be identified.
AI can continue to be part of the creative process, as long as it is transparent.
Blog
