AI film disclosure in 2026: the rules your team needs to know

Article 50 distinguishes the provider of an AI system from the organisation or person using it. Providers of systems generating synthetic media have marking and detectability obligations, subject to the rules’ qualifications. Deployers have separate duties to disclose certain content to people exposed to it.

For image, audio and video, the deployer rule addresses content that constitutes a deepfake. The law also provides an adapted disclosure requirement when such content forms part of an evidently artistic, creative, satirical or fictional work. This is a qualified obligation, not a blanket permission to leave every AI film unlabelled.

The Commission’s guidance considers what viewers would reasonably understand about the content. Its examples explain that standard film effects and background scenes are not automatically deceptive just because AI was involved. The assessment depends on the context and audience as well as the resulting image or sound.

Where a deepfake disclosure is required, the Commission says it must be clear and perceivable. Embedded technical metadata alone does not fulfil that audience disclosure duty. Its guidance also describes a limited later deadline for the marking obligations of certain systems already on the market. That qualification should not be treated as a general postponement of every disclosure rule.

On 10 June 2026, the Commission published a voluntary Code of Practice to help organisations meet the marking and labelling requirements. It provides a practical route for implementing obligations that remain distinct from the voluntary choice to sign the code.

For a production, start with a simple internal record: which supplier or team used which tool, what material or activity it affected, and which delivered version contains it. Keep the approval and disclosure decision with that version. When a trailer is recut or a sequence is replaced, revisit the record.

A legal or distribution review can then work from concrete facts instead of a vague assurance that the film is compliant. This article describes the EU transparency framework as checked on 30 September 2026. A release in several territories also needs advice on the laws, contracts and platform terms that apply to that distribution.

A production can disclose AI use accurately while making no claim that all of its creative work met a human creation standard. If your brief promises human craft, that promise needs its own evidence and review process.

DigiBridge’s strict programme excludes prohibited AI assistance and software acting in place of the contributor. Its verification agent assesses authorised activity alongside device, system and tool checks. Findings retain their reasons and evidence, with an opportunity for the creator to explain or challenge them.

The first confirmed incident holds the affected recording or version. A third distinct confirmed incident makes the project ineligible for directory publication. A model suspicion alone is not a confirmed incident. This gives the programme a defined consequence and a review process, while keeping legal compliance and creative verification separate.