A company commissions a professional photo shoot, receives the images and then uses artificial intelligence to change the backgrounds, create alternative settings and multiply the content intended for social networks. But is it always possible to do it? And what responsibilities does it involve?
Generative artificial intelligence is profoundly changing the field of commercial photography, advertising and visual communication.
Today it only takes a few minutes to take a studio photo of a product, completely replace the background, modify the apparent lighting and set the subject in a scenario that never existed.
A single image can become the basis for numerous advertising content, intended for Instagram, Facebook, LinkedIn, e-commerce sites and digital campaigns.
From a technological point of view, the possibilities are extraordinary. From a professional and legal point of view, however, the issues are more complex.
With the application of the transparency obligations under the European AI Act, starting from 2 August 2026, it has become even more important to distinguish between what is technically possible and what is legally permitted.
When the customer edits a professional photograph with AI
As a professional photographer and videomaker in Rome, I directly experienced this transformation.
I happened to see some photographs made for my clients subsequently processed through artificial intelligence tools, without being involved in subsequent processing.
Product photographs taken with attention to lighting, composition, materials and chromatic rendering have been used to generate new settings.
The original background is deleted, the product is placed in a completely new context and the image is transformed into a different advertising representation.
In this way, starting from a single photo shoot, a company can get numerous variants to feed its communication channels.
It is understandable that this possibility interests companies: it means having more content, speeding up some processes and potentially reducing certain production costs.
But a professional photograph is not simply a file to be transformed without considering rights, agreements and responsibilities.
Can the customer edit a purchased photo with AI?
The answer is: it depends.
In Italy, the protection of photographs is mainly governed by Law 22 April 1941, n. 633, on copyright.
The legislation distinguishes, among other things, photographs with a creative character, which can be protected as photographic works, from so-called simple photographs, subject to a specific regime of related rights.
There is also a particularly important provision for those who take commercial photographs.
Article 88 provides in fact that, for simple photographs of things in the possession of the client, the exclusive right may accrue to the latter, unless otherwise agreed, according to the conditions provided for by law.
This means that it is not correct to argue that any commissioned photograph can only be modified with a new authorization from the photographer.
It is necessary to evaluate the nature of the shot, the contractual relationship, the rights actually transferred and the agreed methods of use.
For this reason, in professional relations, I consider it increasingly important to expressly regulate the possibility of using photographs as a basis for processing with artificial intelligence systems.
A customer should know which uses are included in the photographic commission and which require further arrangements.
In the same way, the photographer should be able to clearly define the value of his production and any limits of use.
AI Act: what changes from August 2, 2026
European Regulation 2024/1689, known as the AI Act, introduces a common regulatory framework for the use of artificial intelligence systems in the European Union.
Among the most interesting provisions for the world of visual communication we find article 50, dedicated to transparency obligations.
From August 2, 2026, specific obligations apply to AI system suppliers and those who use them professionally.
In fact, the regulation distinguishes two fundamental figures:
The provider, i.e. the provider of the artificial intelligence system, is subject to specific technical obligations, including those relating to the recognition of certain artificial content.
The deployer, or the subject who uses an artificial intelligence system under his own authority, outside of personal non-professional use.
In the context of an advertising production, the deployer can be a company, an agency or a professional.
It does not necessarily coincide with the person who materially performs the processing: it also counts who exercises control and responsibility over the use of the system.
Does a photograph with the background modified by AI have to be declared?
This is one of the points on which there is more confusion.
The AI Act does not stipulate that any photograph modified through artificial intelligence must necessarily be accompanied by a visible statement.
Article 50(4) introduces a specific obligation for deployers who generate or manipulate deepfake images, audio and video.
According to the regulation, a deepfake is content generated or manipulated with AI that resembles existing people, objects, places, entities or events and that may appear falsely authentic or truthful.
Let's think of a product photograph.
A company could take a studio shot on a neutral background and then use AI to place the product inside a villa, hotel or commercial environment that has never been photographed.
The final image could suggest the existence of a real situation.
However, not every virtual setting automatically constitutes a deepfake.
It is necessary to evaluate the content, the context, the degree of realism, the possible perception of the public and the conditions provided for by the legislation.
For manifestly artistic, creative, satirical or fictional works, the regulation also provides for special information procedures.
The distinction is especially important for companies that produce catalogs, advertising campaigns and content for social networks.
Who is responsible: photographer, agency or company?
Let's imagine that a company commissions a photo shoot to a professional.
The photographer creates and delivers traditional images, without using generative tools.
Subsequently, the company's marketing department decides to process some photographs with an artificial intelligence system.
In this case, the subsequent use of AI cannot be automatically attributed to the photographer who made the original images.
The position of the company and any external parties involved must be assessed according to the definitions of the AI Act and the control exercised over the system.
The European Commission has made it clear that even a company that uses external collaborators can retain the role of deployer when the system operates under its authority and responsibility.
Therefore, commissioning the production of a content does not mean being automatically exempted from the obligations provided for by the legislation.
AI Act sanctions: up to 15 million euros or 3% of turnover
One of the aspects that deserves more attention concerns the sanctioning system.
Article 99, paragraph 4, of EU Regulation 2024/1689 provides, for certain violations of regulatory obligations, including those relating to Article 50, administrative sanctions that may reach:
EUR 15 million or 3% of the company's total annual turnover for the previous financial year, according to the criteria laid down in the Regulation.
These are maximum limits, not automatic penalties.
For small and medium-sized enterprises, including startups, the lowest limit between the fixed amount and the percentage of turnover applies.
In the concrete determination of the sanction, the circumstances of the violation, its seriousness, the duration and the other criteria provided for by the legislation must also be taken into account.
It is essential to clarify one aspect: replacing the background of a photograph with artificial intelligence does not automatically violate the AI Act and, consequently, does not automatically result in a penalty.
The sanctioning risk arises when specific applicable obligations are violated.
It is also necessary to distinguish these sanctions from any civil consequences deriving from the violation of rights to a photograph.
These are different legal areas and should not be confused.
Is the technical marking of the artificial contents sufficient?
The AI Act also distinguishes the technical obligations of system providers from those that may concern those who use the content.
Article 50(2) governs the marking in machine-readable format of artificial outputs, in accordance with the requirements and exceptions provided for by the legislation.
A technical marking can facilitate the recognition of the artificial origin of an image.
But when the declaration obligation provided for deepfakes also applies, the mere presence of a marking that is not perceptible by the user is not enough to replace the information requested from the public.
The statement must be clear and distinguishable and must be provided, at the latest, at the time of the first exposure to the content.
The legislation also provides for a transitional period…