Legal Recommendations: Model Releases and the Use of Images of People

What does a Model Release actually cover?

A Model Release (MR) gives permission to use specific photos or videos from a session — it does not transfer rights over the person's identity in general. This means that editing, combining, or transforming the original content may be covered, depending on the scope of the specific MR, but generating new images of the same person in contexts not captured in the original is not. For example, face swaps, identity transfers, or placing the model in sensitive contexts (health conditions, religion, politics, sexuality) are generally not permitted under standard MRs and would require specific, informed consent that we do not assume to exist by default, as these fall outside the scope of what the model typically consents to in a standard session release.

Note also that, a person's consent to the use of their image is revocable at any time. This means an MR is never fully irrevocable and a Contributor's MR may cease to be valid if the person depicted withdraws consent.

Does my Model Release need to mention AI uses?

Yes, and this is important. If your MR does not explicitly mention the use of images for AI training, generating variations, or similar advanced uses, Freepik will not apply those uses to your content — even if it is published on the platform. Each MR can be different: some are broader, some more restrictive. Contributors are responsible for ensuring their MRs are valid and sufficient for the uses they enable on the platform. We may request verification of any MR at any time. If the MR cannot be verified for a specific use, that use will not be applied to your content.

Which AI Suite tools can be used with images of real people?

This depends on what the Model Release covers. Here is a general overview:

Regardless of the tool, any output that places a real, identifiable person in a context that could damage their reputation or honor is never permitted — this applies independently of what the Model Release states.

Some of these tools may generate outputs subject to transparency and labeling obligations under the EU AI Act (in particular, art. 50 for deepfakes and AI-generated content depicting real persons). These obligations apply independently of the MR and must be complied with by all users of the AI Suite.

Do AI-generated people require a Model Release?

Generally, no. If a person has been entirely generated by AI and is not based on any real individual, there is no image right to protect and no MR is required. However, if an AI-generated image closely resembles a real, recognizable person, it could still trigger legal protection (image rights, honor, and — where applicable — protection of deceased persons' image rights, which under Spanish law extend up to 80 years post mortem). The relevant criterion is not how the image was created, but whether someone could identify a real person in the result. When in doubt, treat it as if a real person were involved and document the generation process (prompts, base references, model used) to support due diligence in case of a claim.

Even where no MR is needed, other rights may still apply: copyright over base images used in the generation, biometric data protections if training data included identifiable real persons, and trademark or publicity rights if the output reproduces a recognizable iconic image or persona.