What the AI Language in Crew Contracts Means on an Actual Set

The Clauses Are More Specific Than the Discourse

Most conversation about AI on set operates at the level of vibes. The contract language is considerably more precise than that, and it is worth knowing what it actually says if you are working under it or hiring under it.

The core provisions are these. Generative AI use is treated as covered work performed by a human, which means it falls inside the union agreement rather than outside it. No crew member can be required to write a prompt that would displace a fellow crew member. Workers can refuse to be scanned for machine learning purposes. Studios indemnify workers against lawsuits arising from their use of these tools. And training is part of the deal, both for people who need to use the tools and, through cross-training, for people displaced by them.

Notably, crew cannot refuse to use generative AI as a tool. The bargain was not a prohibition. It was an attempt to make sure that using it stays union work.

What That Protects and What It Does Not

This is the crux of the disagreement inside the membership. Supporters argue the language is more evolved than what other guilds secured, because it gets ahead of the pattern where a new technology creates a parallel non-union workforce, which is roughly what happened with visual effects for decades.

Critics, including members of the Art Directors Guild who recommended voting against ratification, argue the protections address the wrong risk. Their position is that the threat is not being replaced mid-job, it is never being hired, and that minimum staffing requirements are the only mechanism that actually addresses that. They also point out that the definition of AI systems is loose enough to cover a lot of proprietary software. Generative tools have already been used to produce commercials with no crew involved at all, which is work that never enters the agreement in the first place.

Both readings are defensible. The agreement ratified by a wide margin, and the objections were substantive rather than reflexive.

The Background Everyone Is Arguing Against

The reason the AI language carries so much weight is that it landed on top of an already severe contraction. When a third of the jobs in the largest market have already gone, any provision that does not directly create hiring reads to a lot of members as insufficient, whatever its legal merits.

What This Means Day to Day

For crew, three practical things. You can decline to be scanned. You cannot be made to prompt a system in a way that takes a colleague''s job, and if you are asked to, that is something your local wants to know about. And if you use these tools as part of covered work, you are indemnified rather than personally exposed.

For producers, the practical thing is that using generative tools in your pipeline does not move that work outside the agreement. If you are on the Low Budget Theatrical Agreement, note that it now carries its own dedicated AI article, so the rules apply at that scale too. Building a workflow first and checking the article afterward is the expensive order to do it in.

Where ABRAM Sits On This

Worth being direct, because a platform that uses AI writing about AI in production should say where the line is.

ABRAM uses AI on the paperwork. Project intake reads a brief and turns it into structured scope. Crew matching narrows a large pool to relevant candidates. Those are scheduling and coordination problems, and the alternative to automating them is not a person being paid to do them, it is a producer doing them badly at eleven at night.

What it does not do is generate the work. There is no version of this where the platform makes the image, replaces the department, or produces the deliverable. The whole premise is that more of the operational overhead should be handled so that the people doing the craft can spend their time on the craft and get paid properly for it.

Reasonable people in this industry disagree about where that line should sit, and the disagreement inside IATSE is evidence that it is genuinely contested rather than settled. The next round of negotiations in 2027 is where the staffing question that this contract did not resolve is likely to get argued again.