Should AI Usage Be Explicitly Disclosed in Movies and TV Shows?

As Netflix reveals that 300 of its titles already feature generative AI, we consider if audiences perhaps deserve better information about AI use in their favorite TV shows, and in movies.

Opinion The first time that a Hollywood movie used the disclaimer ‘No animals were harmed in the making of this movie’ was in 1972, for the canine-heavy American crime comedy film The Doberman Gang. Eight years later, accusations of animal abuse on Michael Cimino’s Heaven’s Gate would standardize the practice.

To the best of my knowledge, the first time that an equivalent notice appeared in a movie, in regard to the use of generative AI, rather than animal welfare, was in the 2024 Hugh Grant horror outing Heretic:

The horror film ‘Heretic’ appears to have been the first to carry a ‘No AI’ disclaimer. Source

Heretic’s directors, Scott Beck and Bryan Woods, stated:

“We have no illusions that when people watch ‘Heretic’ they’re going to go, ‘Wait, did they use generative AI?’” he says. “It doesn’t feel like that at all, but it was important for us to put that out there because we think it’s something people need to start talking about.”

This is a rare standpoint, however. In general, the TV and movie industries currently seem to be waiting out the (they hope) ebbing tide of public discontent around AI usage in their favorite shows and in new movies.

After all, despite the campaigning of CGI dissidents such as Quentin Tarantino and Christopher Nolan, public tolerance of ‘traditional’ computer-generated imagery has, by now, reached a level that can only be defined as ‘fatalist’, or ‘resigned’. The same is expected, apparently, for the use of AI in shows and films.

Hidden Layers of AI

The difference is that the advent of CGI not only put very few people out of work, but instead ballooned the hitherto parochial visual effects industry into a gargantuan media sector employing hundreds of thousands of people around the world. Under this evolution, big ticket VFX movies would transform from much-anticipated tent-pole rarities to (for some viewers) tediously frequent events.

AI, instead, in simultaneously co-opting many of the roles in a single VFX pipeline, is severely and negatively affecting the careers of visual effects workers, and causing VFX and film & TV companies to close ranks and obfuscate AI usage in multiple ways.

For instance, though companies such as DreamWorks may stipulate that their work may not be used to train AI (image below), it’s the undisclosed AI practices of visual effects companies that misrepresents the extent to which AI is used in popular media*.

DreamWorks’ disclaimer in Bad Guys 2 warning against using the movie as training material. Source

The Credit Crunch

All this, because discussions around the use of AI in media output have reignited today, with the disclosure, via a company report and a subsequent interview, that Netflix has used generative AI in around 300 of its titles. Netflix Co-CEO Ted Sarandos stated that The American Experiment docuseries alone features 17 minutes of genAI content.

You would certainly need to be told this: a relatively close check on the 5 minutes+ closing credits of The American Experiment (and another AI-cited Netflix show Glory) reveals zero evidence of AI usage:

Credits relating to multiple VFX companies in ‘The American Experiment’.

The roles cited could have been included in credits at least as far back as the early 1990s. Where’s all the AI?

Since Metaphysic.ai’s pioneering industry breakthrough around 2023 (full disclosure: I worked there for two years), at a time when companies were considerably more vocal about their AI VFX work, AI techniques have quietly become widespread across long-established VFX companies that have since integrated generative AI into their existing workflows.

In this way, AI is creeping into general production with zero fanfare – not least because the ‘telltale’ roles in an AI VFX crew are not generally included in the credits; even a credit for the now-disbanded Metaphysic.ai itself – which should clearly reveal many of the AI-specific roles involved in generating and processing datasets, as well as rendering output via neural networks and other ML methodologies – in fact surfaces only the most generic and ‘traditional’ VFX roles, for the 2024 Alien: Romulus franchise entry:

Credits for Metaphysic for ‘Alien: Romulus’ (2024) – a participation that the author personally witnessed over the course of 18 months. No AI-specific roles appear.

If it were not for the name of the company, there would be no indication of AI usage at all for Alien: Romulus. In the real world, such dedicated work requires workers solely devoted to data-wrangling, training, and numerous other AI-specific tasks; and the presence of multiple staff with the role ‘AI artist’ and similar at the company’s remnant LinkedIn page^† is far less shy of the term. Here, one of the ‘Digital Artists’ credited above for Romulus has an AI-related title instead.

Of course, while credits are longer than they used to be, production companies are not required to exhaustively list everyone, and the Producers’ Guild of America (PGA) issues guidelines to this effect. Nor are VFX credits obliged to name tools used (though the most popular proprietary or named FOSS systems, such as Flame or Houdini, frequently receive a name-check in the credits).

Nonetheless, can it be that hiding AI-related platforms, tools, processes and methodologies from being explicitly named in TV and movie credits is a safeguard against later legal exposure, as the global case against web-scraping evolves? Or else that excluding AI roles and processes from credits allows companies such as Netflix to control the narrative over AI’s incursion with more precision?

Who Cares?

AI evolves far faster than the language that defines and explains it; for instance, the term ‘deepfake’ has changed and extended in meaning considerably since it was used to describe the output of the 2017-era banned software that later evolved into the DeepFaceLab and DeepFaceLive.

I believe that a similar evolution is occurring with the term ‘AI slop‘. This pejorative reference was originally intended to point out the tropes and predictable repetitions of LLM output, as well as the often generic or unconvincing quality of AI image or video generation.

Now, in spite of improvements in quality, the term is becoming a simple moral rejection of AI itself, on political or economic grounds, regardless of the quality of output^††.

As I’ve noted before, the scope for ‘popular resistance’ among consumers may be limited by the B2B nature of the current AI revolution, which, rather than being governed by consumer demand, suffuses machine learning systems into underlying infrastructure, ensuring that most end users engage with AI by default, rather than by choice.

As far as Netflix is concerned, Ted Saranos is relying on consumer indifference to the means by which they get their latest shows and movies; he believes, plainly, that audiences don’t care:

‘I think [audiences] largely care mostly about connecting with the storytelling. I’d say they probably don’t care much about budgets, and arguably not even about the technology to deliver it.’

Perhaps he is right; perhaps no-one ever did go broke underestimating the indifference of mass audiences.

Further, some may argue that the diffusion of AI is so extensive that even a vibe-coded plugin for Houdini would arguably trigger the necessity for an AI disclosure in a movie or TV episode; and that this would lead to the same numbing effect as the EU’s cookie law, GDPR.

Perhaps too, even those who are now ideologically poised against all AI (because they perceive it as an existential threat to themselves or others) will weary of protest; most vegans I know do not forbid themselves from watching movies and TV shows filmed on celluloid, even though these are animal products (as are x-ray investigations).

For some, any amount of AI represents the inkdrop that colors the entire glass, and renders the surrounding product as ‘AI slop’; and the more perfectly AI integrates with its host, the more discolored the host product appears to such a viewer.

To this extent, ‘AI slop’ is not about quality, but rather provenance and solidarity in the face of a perceived threat. Even if just for a ‘transitional’ period prior to enforced normalization, there is an argument for disclosure – even advance disclosure, as in the (AI) mockup below – of AI usage in popular TV and movie output:

Itself depicted by AI (GPT-2), a possible addition to statutory warnings about content in TV shows and movies.

Those who consider that such notices will quickly become redundant might consider that violence, profanity and drug use/abuse are perennial, and probably eternal, and appear with great frequency in TV shows that contain depictions of these domains; yet they are still disclosed routinely, notwithstanding the extent of any repetition.

In much the same way that allergens such as celery are often indicated with emphasis on food packets, it does not seem unreasonable to treat the inclusion of AI likewise – as a choice that the audience can make or not make. Nor, perhaps, should they have to read the entertainment press, or subscribe to company updates in order to find out this information.

* I worked for several years as staff on AI-centric visual effects company Metaphysic.ai, until it was subsumed into VFX behemoth Double Negative (DNEG) as Brahma.ai. Consequently I gained some interesting insights into AI practices in the global VFX industry.

^†You will need to be logged in to see this link.

^†† For this reason I have begun to more explicitly ‘frame’ the article illustrations for my output here as AI-generated, as more and more people appear to be demanding a distinction and disclosure. Prior to this, the AI-generated nature of the images was only embedded in the HTML tags for the pictures.

First published Friday, July 17, 2026. Amended 16:52 EET to remove extra space.