The Three AI Questions Every Producer Should Be Able to Answer

Since 2 August 2026, the EU AI Act’s transparency rules have applied to certain uses of AI-generated and manipulated content. For a film or television producer, the practical starting point is not mastering the whole Act, but being able to answer three questions about each production: what AI was used, whether any resulting content needs to be disclosed, and what rights position the production can support.

Those questions do not cover every legal issue that AI can raise, but they reveal whether you have the information needed to deal with the issues most likely to arise at delivery, during due diligence or when content reaches an audience.

What AI did you use, and where did its outputs go?

Start with the facts. Which tools were used, what tasks were they used for, and did any of their outputs enter the finished programme, its promotional material or another deliverable?

Without that information, later questions about disclosure, copyright, contractual permissions or insurance have to be answered from memory. A light, running record is usually more useful than an investigation after delivery, particularly when AI has been used by several departments and through tools that were not procured centrally.

The record should also be described honestly. It is a dated account of what contributors reported about their AI use, not forensic proof of everything that happened. Even with that limitation, it gives a producer a much stronger basis for review than scattered emails and recollections gathered months later.

Does any content need to be disclosed?

The relevant transparency obligation is found in Article 50 of the EU AI Act (opens in a new tab), which has applied since 2 August 2026.

Where a production company is the deployer of an AI system that generates or manipulates image, audio or video content constituting a deepfake, it must disclose that the content was artificially generated or manipulated. The Act defines a deepfake as content that resembles an existing person, object, place, entity or event and would falsely appear authentic or truthful. The disclosure must be clear and distinguishable and provided, at the latest, when the audience is first exposed to the content.

The treatment of creative work is more flexible, but the duty does not disappear. Where the content forms part of an evidently artistic, creative, satirical, fictional or similar work, disclosure may be made in an appropriate way that does not hamper the display or enjoyment of the work.

Not every use of generative AI in a production will meet the definition of a deepfake, and the answer depends on the resulting content and its context rather than the tool name alone. The European Commission’s Article 50 guidelines (opens in a new tab) provide further examples of what falls inside and outside the rules.

This is why the first question matters: you cannot assess a disclosure duty unless you know what was generated or altered and where it appeared.

What rights position can you support?

“Do we own it?” is often treated as a yes-or-no question, but AI makes the answer more dependent on the particular workflow.

Under EU copyright law, protection requires an original work that expresses its author’s free and creative choices. How that test applies to a particular AI-assisted or AI-generated asset will depend on the human contribution visible in the finished work, and the boundary is not yet governed by a simple EU-wide rule for generative AI. Tool terms, permissions for source material, performers’ rights and national law may also affect what a producer can use, assign or warrant.

A contemporaneous record can help here if it captures the human contribution: who directed the work, what choices they made, how outputs were selected or rejected, and what further editing took place. It may support an assessment of authorship and chain of title, but it does not create copyright where the legal requirements are absent, cure an unauthorised input or guarantee that an output is free from third-party claims.

The useful question, therefore, is not simply whether the production owns everything it made. It is whether the producer has enough information to identify the relevant rights and support the position taken at delivery.

Why these three questions?

These are organising questions, not a complete compliance test. Depending on the workflow, a production may also need to consider personal data, confidential material, performers’ likenesses and voices, employment-related systems, commissioning terms and the conditions imposed by individual AI providers.

The three questions still provide a useful first pass because they establish what happened, identify content that may require disclosure, and expose rights issues that need closer review. A production record does not make those legal questions disappear. It shows you which questions actually need to be answered.

For a closer look at the disclosure duty, read what changed on 2 August 2026.