Jun 4, 2024 - Technology
Human Intelligence

The fight to keep documentaries real

a photo illustration of Stephanie Jenkins and Rachel Antell of Archival Producers Alliance on a circuitboard background

Archival producers Stephanie Jenkins and Rachel Antell. Photo Illustration: Axios Visuals; Photos: Josh Goleman, Natasha Benjamin

AI will wreck the documentary tradition unless filmmakers track and disclose their use of generative AI, according to a group of award-winning archival producers who are fighting for new standards.

Why it matters: With no laws in place or professional codes governing how documentarians use AI, public trust in the visual record of history will erode, the three women who co-founded the Archival Producers Alliance (APA) argue.

Catch up quick: Archival producers research and manage footage, photographs, audio and ephemera for use in documentary films.

  • Rachel Antell, one of the APA co-founders, told Axios she was working on a documentary last year, soon after ChatGPT's debut, when the filmmakers showed her images they'd made with AI that they planned to use alongside real archival photos.
  • "We were very alarmed," Antell recalled. "Are you going to tell the audience this is not a real image of this historical figure?'"
  • Antell and her colleague Jen Petrucelli joined forces with Stephanie Jenkins, another archival producer, to start the APA — with its first initiative aimed at sorting out the use of generative AI in documentary.

The big picture: Archival producers are in the trenches of the fight to maintain accuracy in documentary film.

  • The APA, with help from a global group of volunteer experts, has proposed a set of guidelines designed to recognize what genAI brings to filmmaking while also addressing the risks.
  • "Things are moving so fast and filmmakers want to use this technology," Antell said. "And we think there are a lot of risks to that."
  • The guidelines aim to "preserve the ethical nature of documentary and preserve the field of documentary as a genre," she said.

Between the lines: The long-term problem is that documentaries build on other documentaries, and if filmmakers can't trust the authenticity of their predecessors' work, the whole enterprise collapses.

  • "Generated material presented as 'real' in one film will be passed along — on the internet, in other films — and is in danger of forever muddying the historical record," the APA wrote in late 2023 in an open letter signed by over 100 people in the field.
  • GenAI is of particular risk in documentary film, the APA claims in the letter, because "the commingling of real and unreal taints all of it; if neither images or audio can be believed, then the nonfiction genre is hopelessly compromised."
  • The APA guidelines recommend production transparency practices — like tracking genAI use with a cue sheet as you would music, including details about software versions and prompts used.
  • They also suggest adding watermarks to generated images, or mentioning the process in the film itself.

Yes, but: Antell and Jenkins said they don't want to fence filmmakers off from ever using genAI.

  • " We recognize there may be good reasons for documentarians to use AI-generated media: to bring to life the stories of people who are missing from history, or to take viewers to a time or place without adequate visual representation," the APA's open letter says.

Documentary film and photography have always raised thorny questions about truth and authenticity.

  • Case in point: In the 19th century, pioneering war photographers moved corpses around the battlefield in order to tell a particular story.

But using genAI in documentary film is different, Antell and Jenkins said.

  • "The act of humans bearing witness is an integral part of the documentary genre. It can't be fully separated out," Antell said.
  • Documentaries, Jenkins says, are "documents. They're a filmmaker going through and saying, 'Look at all this research I've pulled together, I'm creating a new document about this subject.'"
  • GenAI tools, the producers argue, can't replace this because they're machines, Jenkins says: "Machines [aren't] born. They don't change in the way humans do and they don't die."

What we're watching: The APA plans to release its AI guidelines in September.

Go deeper