When AI enters the edit bay: the Soderbergh question no one's asking
Steven Soderbergh fed a rough cut of his new documentary to Meta’s MovieGen AI and let it suggest edits. The trades wrote it up as a workflow innovation. The tech blogs praised the collaboration. What stays with me is the woman in rural Kentucky whose face appears in that footage. Did anyone ask her whether she consented to an algorithm analysing her expressions and deciding which moments of her life make better cinema?
I am not panicking about AI tools as such. I have watched editors use algorithmic colour matching for years without any ethical alarm going off. This is a different thing. Point AI at documentary footage and you are asking a black-box system to make aesthetic judgments about real people’s lives, people who never agreed to have their vulnerability run through a statistical model trained on God knows what. Soderbergh’s experiment shows off a powerful editing tool. It also shows how quickly consent gets skipped once the technology becomes exciting.
Documentary subjects live in a strange legal twilight. Yes, they sign releases. But those releases were drafted when “editing” meant a human watching footage and making cuts. The implied contract ran between subject and filmmaker, a relationship between people, however lopsided. Now there is a third party in the room who never shakes hands, never builds rapport, never sits with the discomfort of deciding whether someone’s tears belong in the final cut. The algorithm feels nothing when it flags an “emotionally resonant moment.” That is exactly the problem.
I have spent enough hours in post-production suites to know that editors carry their subjects with them. They argue over whether showing a breakdown serves the story or just serves spectacle. They weigh context the camera never caught: what happened after the recording stopped, what the subject said off the record, whether this scene might set back a fragile recovery. All of that happens through human judgment, informed by human conscience. Hand even part of it to AI and you are outsourcing ethics to efficiency.
The Soderbergh case is harder still because some of his subjects have died. The documentary covers his parents’ final years. Whatever consent conversations took place, they took place with people who cannot now respond to how their images are being processed. Film ethics has always wrestled with posthumous representation, but at least a human editor could hold themselves accountable to memory, to stated wishes, to the weight of depicting the dead. An AI system has no such tether. It processes deceased subjects with the same cold objectivity it brings to everything else.
None of this is hypothetical. Productions are already using AI tools in assembly edits, and the savings are real. A rough cut that might take a junior editor two weeks now takes two days. Yet I have never seen a budget line for “additional subject consent for algorithmic processing.” I have never seen a release form that mentions machine learning. We are moving forward on the assumption that consent given for traditional documentary work transfers automatically to whatever tools we invent next. That assumption is legally shaky. Ethically, it is bankrupt.
The European productions I work with are starting to notice. GDPR provides at least some friction; you cannot simply push footage of EU citizens into a US-based AI without thinking hard about data sovereignty. Most of the documentary world, though, is not asking these questions yet. We are still in the “wow, look what it can do” phase. I have been in that phase before, with other technologies that looked purely beneficial right up until someone stopped and asked who was carrying the unexamined costs.
What makes this urgent is that documentary subjects are often vulnerable almost by definition. We film people in crisis, in poverty, in grief. They agree to take part because they trust the filmmaker, not because they trust the whole technological supply chain behind the film. When someone shares their story with a director they have spent months building a relationship with, they are not picturing that footage being parsed by the same class of tools that power TikTok recommendations. The betrayal is not intentional. It is real all the same.
Some will say I am inventing a problem. AI editing tools are just faster scissors, the argument goes, and subjects consent to “the film” whatever the process. But process shapes outcome. An algorithm trained on commercial cinema will flag different moments than a human editor working from documentary ethics. It does not know the difference between a useful revelation and an exploitative one. It cannot judge whether a particular cut breaks the spirit of what someone agreed to. It optimises for engagement metrics that were never on the table during the release conversation.
1. Documentary consent forms must explicitly address AI processing of footage and give subjects the option to exclude their material from algorithmic editing tools.
2. Productions using AI editing assistance need dedicated review by someone who knows the subjects and can advocate for their dignity, not just for efficiency.
3. Posthumous subjects require even stricter protocols, with family consultation before any AI-assisted editing of deceased persons’ footage.
Soderbergh’s experiment will probably produce a good film. He is a thoughtful filmmaker with decades of craft behind him. But the trades will not cover the consent conversations that never happened, the questions nobody thought to ask because the technology moved faster than our ethics. That gap is where documentary subjects always end up paying. We can close it, or we can keep tuning our workflows and pretending process does not matter. I know which choice lets me sleep better.
Todd Max Carey, Seven Journeys