AI-generated content in film: the legal aspects
Artificial intelligence is playing a growing role in film productions, from generating scripts and visual effects to creating background music. This raises new legal questions about copyright, clearances and liability.
Suppose you make a short film and use AI for the script, the visual effects and the music. The production is finished, the distributor is interested, and only then does the real work begin. Because who holds the copyright in what the AI made? And how do you demonstrate to an errors and omissions insurer that all the rights are in order?
These are no longer hypothetical questions. As film law lawyers we see them come up more and more often in practice, and the answers are anything but self-evident.
Copyright in AI content: who owns what the AI makes?
The core question is who holds the copyright in content generated by AI. Dutch copyright law requires a human author. A wholly autonomously generated AI script or AI image may not qualify for copyright protection. Where AI is deployed as a tool under a person’s creative direction, copyright can indeed arise in that person. But it is rarely that simple. The boundaries have not yet crystallised.
In practice AI is almost always deployed as a tool: you give prompts, make deliberate creative choices, select and edit. The greater the human creative input, the greater the chance of copyright protection arising in the person. Exactly where the line falls is still very much in development. Courts in Europe and the United States are already grappling with it, and the outcomes differ.
What does this mean for you as a maker? Document your working process. Which creative decisions did you make? Which AI tool did you use, and how? That documentation can make the difference if a dispute arises later.
AI in your chain of title: how do you record it?
Distributors and broadcasters ask as a matter of course for a watertight chain of title: an overview of who acquired and transferred which rights. For traditional content that is a familiar exercise. For AI-generated content the rules of the game have not yet crystallised.
Insurers writing errors and omissions policies are wrestling with the same question. To obtain errors and omissions insurance you have to be able to demonstrate that all the rights are in order. Some still refuse cover for AI elements, others impose additional requirements. Anyone who fails to take these questions into account during the production phase runs the risk that distribution gets stuck later.
Practical advice: record during the production phase how AI was used:
- which human creative choices were made
- what licence terms the tools used impose
- which rights you claim as a user.
Not every AI platform automatically gives you the rights in the output, so check the terms of use as well.
Rights in AI training data
AI systems are trained on existing content: films, music, photographs, texts. Where that training data is protected by copyright, the question arises whether use of the AI output does not indirectly amount to infringement. This plays out particularly with AI-generated music and visual effects that bear a recognisable resemblance to existing work. Several court cases on this are running in the United States and Europe. The outcome may have direct consequences for content now in production.
Collective agreements and stakeholder arrangements on AI
Following the Hollywood strikes of 2023, American collective agreements contain new provisions on the use of AI. In Europe too, attention from unions and guilds is growing. In collaborations with screenwriters, actors and other creatives, take account of these arrangements and record expressly in contracts how AI is or is not used.