DEEPFAKE
DEFINITION
A deepfake is synthetic audio or video generated by AI to depict a real person saying or doing something they did not, typically produced by training a model on genuine footage or recordings of that person.
WHAT IT MEANS IN PRACTICE
A deepfake sits at the intersection of copyright, likeness rights, and increasingly dedicated AI regulation, because the output usually infringes no existing copyrighted work directly but still exploits a real person's identity without consent.
The EU AI Act now requires clear labelling of AI-generated or manipulated audiovisual content that depicts real people, under its transparency obligations in Article 50. The US has no equivalent federal labelling law, though several states have passed narrower deepfake statutes aimed mainly at election material and non-consensual intimate imagery.
For makers, the practical risk runs in both directions: being the unwitting subject of a deepfake, and being asked to build tools or content that could produce one. Contracts with AI vendors increasingly need explicit deepfake and consent clauses.
AN EXAMPLE
A political consultancy creates a synthetic video of a mayoral candidate appearing to make a statement she never made, then distributes it without any AI label. In the EU this breaches the Article 50 disclosure duty; in the US, the consultancy may face liability only if a specific state deepfake statute applies.
WATCH OUT
Labelling a deepfake does not automatically make it lawful, since likeness and defamation claims survive a disclosure notice.
LAST REVIEWED