Here is a question nobody is ready for.
A procurement officer at a Saudi government entity looks at the AI film you delivered six months ago. She points at a face in frame four hundred — a background character, on screen for two seconds, generated not cast.
Where did that face come from? What was the model trained on? Is there a real person who has a claim on it? Who approved the shot? Can you show me?
Most AI studios working in this region today cannot answer a single one of those questions. Not because they did something wrong. Because they never kept the record.
The warning already came, and it came from inside the industry
On 12 October, Arab News published an interview with the Saudi director Ali Al-Kalthami. It deserves more attention than it got.
Al-Kalthami is not an AI sceptic. He says he uses it himself — during development, to test ideas, to see problems before a shoot. "For me, it is a tool, not a goal." He is explicitly dismissive of reflexive fear: "Putting up barriers and fearing it outright will not be the answer."
That is what makes the rest of what he said serious.
He draws a hard line at AI-generated features made for commercial release where the purpose is to replace people and cut costs. And then:
"There has to be legislation to regulate this. It has already started in Europe and America, and I think it is coming to the Arab world and to Saudi Arabia too. The goal is to protect the cultural sector and the creative economy and support their healthy growth."— Ali Al-Kalthami, Arab News
A practitioner who uses the tools is asking for rules about the tools. That is not a culture-war position. That is an industry maturing, and it is how every regulated creative sector got its rules — from inside.
What the industry currently offers instead of an answer
I went looking at what my competitors say about this. The pattern is consistent and it is not encouraging.
Detroit in Paris — the most credentialed luxury AI production house in Europe, working with Louis Vuitton, Dior Parfums, Hennessy, Guerlain — states that its pipelines are "secure, traceable and aligned with brand, legal and ethical standards." That is a good sentence. It is a sentence.
PROMPTR in Berlin lists "secure production pipelines and legal clarity" among its reasons to choose them. One bullet, in a list of four.
Xawiya in Dubai went furthest. In June they launched two named products through Campaign Middle East: ANA, for IP and talent rights, and ARCC, for AI readiness and compliance. Genuinely the right idea, and first to it.
I have now checked their website on three separate occasions over two months. Neither product appears on it. Not on the services page, not in the navigation, nowhere. The offering exists in a trade press announcement from June and, as of this month, in their sales database profile. It does not exist anywhere a buyer would actually look.
I am not scoring points. I think Xawiya identified the opportunity before anyone else did and then could not ship it, which is a thing that happens to all of us. But it tells you how the whole category is positioned right now:
Everybody has the sentence. Nobody has the file.
Why the Gulf gets here first
If you work in Los Angeles you can treat this as a 2028 problem. In the Gulf you cannot, for three structural reasons.
Public money attracts scrutiny earlier. A meaningful share of ambitious production work in Saudi runs through government entities, Vision 2030 programmes, and quasi-public developers. Public budgets come with audit expectations that private brand budgets do not. The question gets asked sooner and it gets asked formally.
Cultural representation raises the stakes on likeness. Generating a Saudi family, a Khaleeji wedding, a scene in a mosque courtyard — these are not neutral acts. Getting the clothing wrong is an embarrassment. Getting a face wrong, or generating something that resembles a real identifiable person without any record of where it came from, is a different category of problem.
The regional conversation has already started. Al-Kalthami is one data point. Saudi designating 2026 as a year focused on artificial intelligence is another. Regulation tends to arrive faster in markets where the state is an active participant in the sector rather than a distant referee.
My estimate is eighteen months before provenance language starts appearing in RFP boilerplate for government-adjacent work in the Kingdom. I may be wrong on the timing. I am not wrong on the direction.
Four questions every buyer should be asking right now
You do not need to wait for regulation to protect yourself. If you are commissioning AI production, ask these four before you sign. They cost nothing and they will sort the field quickly.
1. Who directed each shot, and can you show me the approval?
Not "we have a director." The record. If a named person signed off on frame four hundred, there should be evidence of it that was created at the time, not reconstructed afterwards.
2. Where did every likeness come from, and what consent covers it?
For any face that reads as a specific person — a presenter, a spokesperson, a recurring character — there should be a documented source and a documented permission. For generated background characters, there should be a stated method and a reason you can repeat in public.
3. What did you discard, and why?
This one surprises people, and it is the most revealing. A studio that cannot show you what it rejected either did not reject much, or did not keep track. Both tell you something about how the work was actually made.
4. If a regulator asks in two years, can you reproduce this answer?
Not "would you try." Can you. Is the record stored, is it structured, does it survive a staff change.
If a studio cannot answer these, that is not a technical shortcoming. Records are cheap. It is a discipline shortcoming, and discipline is what you are paying a studio for.
The uncomfortable part for studios
I will finish on our side of the table rather than the buyer's, because it would be dishonest to write this as though it costs us nothing.
Keeping this record is work. It means your pipeline has to generate documentation as a byproduct rather than as a report someone writes at the end. It means storing generations you would rather forget. It means the discards are visible, which means the failures are visible.
It also means you cannot quietly reuse a look you found somewhere, or let a junior artist's shortcut through to delivery, because the record would show it.
That is the actual reason the category has sentences instead of files. The sentence is free and the file is not.
But I would rather build that discipline now, while it is a differentiator, than retrofit it in eighteen months when it is a requirement and everyone is doing it badly at once.
We have done it once — fully, publicly, on a film where the subject matter demanded it. I will write about what that actually cost us next.
Sources & Notes
Ali Al-Kalthami interview, Arab News, 12 October 2026 (first published in Asharq). Competitor positioning drawn from publicly available company websites and trade press, verified October 2026.