GenAI Best Practices - page test

Opening Statement

We are the Archival Producers Alliance (APA), a group of experienced documentary filmmakers
using our collective knowledge to influence policy and effect change within the film industry. Our
organization includes hundreds of producers and researchers from across the United States and
around the world.
As archival producers, we are on the front lines of preserving the integrity and accuracy of primary
source material used in documentary films. We advise and collaborate with directors, producers,
and editors in researching, sourcing, licensing, and verifying all forms of audio-visual archival
materials. We spend our time faced with the power—and the complications—of representing the
past. A duty to truthfulness is an enduring principle of our work.
We have developed the following guidelines to reaffirm the journalistic values that the
documentary community has long held. While the exact definition of “documentary” will always be
debated, we find G. Roy Levin’s characterization, from his text Documentary Explorations, useful—
that the genre has an “inherent obligation to reality.” A guiding principle of our craft has been
documentary film’s implicit promise to the audience that what is presented as real is in fact real—
even though that reality will, no doubt, have a point of view. There is inherent worth in preserving
the fact-based nature of documentary so that it will continue to be a trusted cultural resource.
Further, we contend that human creativity holds unique value, and that the act of humans “bearing
witness” is an integral part of the documentary genre.
The APA acknowledges and welcomes many of the opportunities that Machine Learning and
Generative Artificial Intelligence (GenAI) bring to filmmaking—and also realizes that this new
technology comes with risks that need to be addressed. For the purposes of these guidelines,
“GenAI” refers to a type of artificial intelligence capable of creating images, video, audio, or other
media in response to prompts from a user. We distinguish GenAI from other Machine Learning, or
other AI, that might be used in production workflows for ideation, transcription, logging, etc. As

additional clarification, this document is not concerned with the minor alteration of existing audio-
visual assets that GenAI can facilitate, such as retouching, restoration, or upres-ing. Rather, it is

meant to address the creation of new materials, as well as alterations to existing primary source
materials that change their meaning in ways that could mislead the audience. As they have always
done, filmmakers must continue to assess where this line lies in each instance.
The possibilities of GenAI are limitless—but there are some burdens only filmmakers can carry. We
intend for these guidelines to help filmmakers, funders, and distributors establish Best Practices
for use of GenAI within documentary film. We propose that the following, overarching principles
guide that work:
1. Value of Primary Sources
2. Transparency
3. Legal Considerations
4. Ethical Considerations of Creating Human Simulations

Guidelines

1. Value of Primary Sources
The APA affirms the irreplaceable value of primary sources. Authentic audio-visual records are
created by humans, at a specific moment in time, contemporaneous with the events they portray.
When used in documentaries, they allow viewers to immerse themselves in a moment in time and
invite the opportunity to experience what someone from the era recorded of that moment, from
their perspective, and with a particular purpose.
As archival producers, we are always aware that the Archive may be biased and/or problematic in
a variety of ways. However, when using primary sources, the owner, context, and intent can be
known, wrestled with, and debated—all key parts of building an understanding of our shared

history. By contrast, GenAI software often pulls from multiple unknown sources, creating realistic-
seeming materials that can mislead audiences. GenAI materials are elements with perceived

authority, yet no accountability of authorship.
When choosing to use GenAI in lieu of primary sources, we recommend considering the following:
A) GenAI and the Historical Record: Synthetic material that is indistinguishable from primary
sources risks being passed along—on the internet, in educational materials, in other films—and is
in danger of forever muddying the historical record.
B) Algorithmic Bias in Synthetic Media: We advise filmmakers to consider the role played by
algorithmic bias when using GenAI to create media—both in reinforcing stereotypes and in
overcorrecting to combat them. GenAI encodes every bias of its dataset. Additionally, most GenAI

models draw from an incomplete version of the historical record, as the majority of physical audio-
visual archives still remain undigitized.

C) GenAI Alterations to Primary Sources: We encourage filmmakers to consider maintaining the
original form and/or medium of primary source material (i.e. still photo vs. moving image, black &
white vs. color, aspect ratio, etc.), or to alert the audience if they do otherwise. The original format
of a primary source can often convey information about the time and place in which it was created.
Using GenAI to alter a primary source may mislead the audience. For example, converting a
photograph to a moving image could imply that a moving image camera was at an event when it
was not. Or altering the aspect ratio of a photograph using generative fill may introduce new
people, buildings, landscapes, or even empty space, to an image that could change its context
and meaning.
D) Re-enactment and Re-creations: Filmmakers have long wrestled with how to maintain
allegiance to the truth while utilizing all the creative tools at their disposal—including
re-enactment and re-creation. We believe the questions posed by GenAI are on a continuum with
these, but that GenAI creations present a risk of greater magnitude because they often require
little time and expense to produce. As such, we encourage filmmakers to treat image generation
with the same intentionality, and the same care for accuracy and sensitivity, as they would a
traditional re-creation, so that human discernment remains central to the process.

2. Transparency
We call on the industry to hold itself accountable by pursuing a transparent production process
when using GenAI, which we break into two categories.
A) Inward: By maintaining transparency inwardly (with the production teams, legal counsel,
insurance companies, distributors, streamers, and subjects) filmmakers will mitigate provenance
questions, ethical concerns, and legal risks.
B) Outward: By maintaining transparency outwardly (with audiences) filmmakers will
safeguard viewers’ trust—guiding them in their understanding of what is authentic and what is
synthetic media within the context of a given film.
Inward Transparency—Production Team
We encourage real-time communication within production teams so all are aware of what is and is
not GenAI material. Production team members should be aware of the proliferation of GenAI
materials on the internet and even in some archives (such as “deepfake” videos, doctored photos,
etc.), and take care not to inadvertently bring them into a project. When filmmakers do choose to
use synthetic media during the production process, we suggest adding a temporary watermark to
those materials to avoid confusion about what is GenAI and what is not.
To document the provenance of GenAI materials used, we recommend production teams produce
cue sheets for all GenAI elements—similar to the deliverables created for music and archival
materials. For each GenAI element, we suggest cue sheets record:
Prompts used, software version (including Terms & Conditions), and the date created (for all
GenAI used that was created by the production team)
Any reference material used (i.e. audio recording, image, etc.), and the copyright status of that
reference (for all GenAI used that was created by the production team)
Descriptions and timecodes for where GenAI materials appear in the production (for all GenAI
used, whether created by the production team or pulled from a third-party source)
Outward Transparency—the Audience
We strongly advise filmmakers to alert and make clear to audiences their use of GenAI. There
should be clarity for the audience about what is and is not synthetic media. We recognize that
different types of transparency may be needed, depending on context and the audience’s
understanding of what is happening on screen. Suggested methods for doing so include:
Lower thirds, bugs, or other types of watermarking on the material itself
Visual vocabulary that alerts the audience to GenAI use, such as a unique frame around the
material, change of aspect ratio, colorization, etc.
Having a character or narrator acknowledge the use of AI in voice-over or on screen

Top of show and/or end of show language, keeping in mind that films are often excerpted—for

educational purposes, publicity, or otherwise—and viewers may not “tune in” to the beginning
or end of a given program. Therefore, we suggest this not be the only means of transparency

Acknowledging the use of GenAI in promotional materials, such as trailers and press releases

We encourage filmmakers to be extra diligent when GenAI is used to do the following—as the risk
for sowing confusion is high:

Make a real person say or do something they did not say or do (i.e., human simulations, a

technique sometimes referred to as a “deepfake,” see section 4)

Alter or create footage, photos, or audio of a real event or place
Generate a realistic-seeming historical scene that did not actually occur

Additionally, GenAI is a new type of media that should be included in end credits in a manner
similar to archival footage, music, etc. These credits might include:

GenAI software or tools used, including which version of software
GenAI creator, such as artist or writer of prompts
GenAI companies employed

3. Considerations
In order to mitigate risk when using GenAI in a documentary, we encourage filmmakers to
undertake a thorough legal review, allowing adequate time and budget for this complex and
important step. Lack of transparency with legal counsel and E&O insurers could to lead to
incomplete or inaccurate advice resulting in additional legal risk for productions (see section 2).
We strongly recommend working with attorneys well-versed in, and up-to-date on, the rapidly
changing legal landscape around GenAI, including (but not limited to) the following areas of
concern:

The multi-layered rights issues for production-created as well as third-party AI-generated materials (e.g., Terms & Conditions of rights-holder licenses, Terms & Conditions of AI software used, etc.)
Requirements of union contracts (including, but not restricted to, those of DGA, WGA, IATSE, and SAG-AFTRA)
Misappropriation of a person’s voice, likeness, or name when using GenAI materials (i.e. Rights of Publicity) and the attendant risk of misrepresentation
Copyright protections, or lack thereof, for AI-created materials that lack sufficient human authorship
GenAI’s potential for misappropriating existing copyrighted materials
The use of GenAI technology exposing the production to risk that is not covered by E&O insurance policy
Suggested disclosure language for indication when GenAI materials are used (i.e. lower thirds, top- or end-of-show, etc.)
Relevance of international laws as they apply to GenAI use

4. Ethical Considerations of Creating Human Simulations:
When using GenAI technology to alter or manipulate the face, body, or voice of a real person
(i.e.“deepfake”), we encourage filmmakers to take extra caution and to consider the following:

Whether a search has been reasonably pursued to find primary source materials which depict

the individual at the particular place and time the production aims to portray

Acquiring consent, when appropriate and relevant, of individuals or constituencies whose

likeness, images, and/or voices are being simulated; and considering ways to take extra care
when a subject or affected individual is unable to give informed consent

Potential cultural sensitivities of using synthetic media to create a person’s face, body, or voice
Whether the use of human simulations in the documentary could adversely affect the historical

record (See section 1)

Whether the use of human simulation is transparent to the filmmaking team and the audience

(see section 2)

Conclusion

As it has always done, documentary film is going to change in ways that stretch filmmakers and
audiences alike. We at the APA are excited by the possibilities that emerging technologies will
bring—especially for stories that have been overlooked, purposefully suppressed, or not recorded
in any fashion. GenAI technology can also help protect the identities of people for whom their
appearance in a documentary presents a risk. Far from being diminished by the challenges posed
by GenAI, there is great potential to enhance documentaries of all kinds by responsibly harnessing
this new technology. That said, we reaffirm the value of human labor and discernment in the
production process.
While these guidelines focus specifically on documentary film, many of the ideas are applicable to
other media that employ archival materials—including narrative films, podcasts, commercials,
museum exhibitions, and more.
We created these guidelines as filmmakers passionate about maintaining the impact of the work
we do. We hope this document and our organization will bring thoughtfulness and intentionality to
the fast-approaching future. We believe that with the integration of standards such as these,
documentary programming will continue to be an engaging, reliable, and most of all, trusted form
of audio-visual storytelling that records human history and expresses human experience.

About these Guidelines

These guidelines are not intended to constitute legal advice; rather, they are provided for
informational and discussion purposes only. If filmmaking teams have a specific legal question
about anything covered in these guidelines, they should consult a legal team. The views expressed
in these guidelines are those of the Archival Producers Alliance only, and not those of any
individual member or his/her/their respective employer.
You may use, reproduce, adapt, and copy relevant provisions of these guidelines provided that you
do not alter or misrepresent the substantive meaning of the guidelines, and provided that you
credit the Archival Producers Alliance in the following form: “Guidelines copyrighted by the
Archival Producers Alliance, and used by permission.”

About the Archival Producers Alliance

Co-Directors Rachel Antell, Stephanie Jenkins, and Jennifer Petrucelli first came together in the
summer of 2023 to discuss concerns about GenAI in documentary film. We began facilitating
meetings as a way to create community among independent archival producers and to develop
best practices around our unique role within the film industry. The enthusiasm from these meetings
encouraged us to formalize the group as the Archival Producers Alliance. Members volunteered to
join various committees to grow the organization, the IDA came on as fiscal sponsor, and we
collectively wrote a set of guidelines around the use of Generative AI in documentary film. Beyond
our GenAI initiative, the APA advocates for the accessibility of undigitized archives, educates the
larger film community about the craft, and works to make archival materials more affordable for
independent filmmakers. The organization received a seed grant from Nancy Blachman, and launch
funding for the AI initiative from the Jonathan Logan Family Foundation.

Methodology

This document is the result of organizing among dedicated archival producers. In virtual meetings
starting in the summer of 2023, the newly-formed Archival Producers Alliance facilitated
discussions about GenAI entering our field. Our members expressed growing unease over the
expanding use of GenAI in documentary films to create ‘fake archival’ materials, such as
photorealistic images that are indistinguishable from primary sources. We shared our experiences
and concerns about the implications for our work, the documentary genre, and the industry at

large—and decided it was time to speak out. Over successive virtual meetings, the group co-
wrote and released an Open Letter, which was signed by over 100 professionals. The Hollywood

Reporter published an article about the APA in November 2023 and printed the letter in its
entirety.
The response was overwhelmingly positive, as we found filmmakers and executives alike hungry for
guidance in the use of this new, powerful, creative tool. An invitation from the International
Documentary Association to present a draft of best practices guidelines at their Getting Real
Conference in April 2024 provided us with a timeline.

From January through March 2024, the following committee of APA members met weekly to
research, discuss, and craft a draft of the guidelines: Rachel Antell, Steve Bergson, Eugen Bräunig,
Bianca Cervantes, Lea Donovan, Christine Fall, Benjamin Field, Stephanie Jenkins, Meaghan
Kelley, Clare Marash, Andrew Mayz, Debra McClutchy, Michelle Moy, Jennifer Petrucelli, Kenn
Rabin, Nacho Ruiz, and Colin Weatherby.
Highlights of the process included:
Researching existing guidelines from PBS, BBC, PACT, WGA, SAG, President Biden’s
Executive Order on the Use of Artificial Intelligence, as well as what streamers have stated
publicly on the topic
Meeting with AI scholars Marie desJardins (formerly of SRI International, University of
Maryland Baltimore County, and Simmons University) and Robert Mahari (MIT Media Lab) to
get their insights on the capabilities of the technology, digital forensic practices, and the state
of AI detection software
Engaging filmmakers John Haptas and Kristine Samuelson, professor emerita at Stanford
University’s Documentary Film Program, to explore definitions of documentary useful to our
guidelines
Speaking at several events and panels, including at the Producers Guild of America, the
Television Academy, FootageFest, Princeton University, and the Big Sky Film Festival
Inviting an array of experts to speak at meetings of the full APA membership:
Patricia Aufderheide, co-facilitator of the Documentary Filmmakers Statement of Best
Practices in Fair Use—which was created by five national filmmaker organizations—spoke
with us about the methodology they employed in crafting their standards
Reuben Hamlyn, co-director of Another Body, presented to the group regarding the ethical
and technical considerations he navigated when using GenAI in his film
Benjamin Field, a consultant on the use of synthetic media in non-fiction films, updated our
group on his work around GenAI in the UK and the behind-the-scenes conversations
currently happening with archives on this topic.
Following this period of intensive research, the committee co-wrote sections of the guidelines in
small groups, coming together frequently to discuss ideas, share language, and edit. An initial
draft was co-edited by APA Co-Directors Rachel Antell, Stephanie Jenkins, and Jennifer
Petrucelli. We then solicited input from the APA membership (approximately 230 people at that
time), and synthesized their observations before sending a subsequent draft to industry leaders for
input. Readers included the experts listed above as well as: Dr. Allyson Slater (technologist, and
data scientist), Matt Dellinger (author, archivist, and AI technologist), Sophie Compton (filmmaker,
co-director of Another Body, and ethical AI advocate), and Nisha Burton (filmmaker and emerging
technology consultant).
Abby Sun (IDA Director of Artist Programs) and Meghan Monsour (Getting Real Conference
Programmer) were instrumental in creating a breakout session at the Getting Real Conference
where Rachel Antell, Stephanie Jenkins and Jennifer Petrucelli presented the first public draft. The
following field experts responded publicly at the panel: Shana Swanson (Supervising Producer at
ITVS), Jon-Sesrie Goff (Program Officer at the Ford Foundation), and Elizabeth Woodward
(Documentary Producer at WILLA).

In the weeks following Getting Real, APA members solicited further input from a wide array of
directors, producers, distributors, lawyers, E&O insurers, and professionals in Standards &
Practices to create this final draft. They received additional feedback from community members at
their presentations at DOC NYC PRO, Sunnyside of the DOC, and the D-Word.
Highlights of this process include receiving feedback and revisions from the following:

Legal counsel, including Loralee Sundra and Dale Cohen (UCLA Law Clinic); Cameron Stracher

(Stracher Law); Chris Perez (Donaldson, Callif & Perez); Steven Beer and Jessica Hidalgo
(Lewis Brisbois Bisgaard & Smith, LLP)

Standards & Practices executives Talia Rosen and Marcia Apperson (PBS)

E&O insurers Evynne Grover and Daniella Granata (QBE North America)

Documentary funders Jenny Raskin and Kelsey Koenig (Impact Partners)
Documentary producers/directors including Brian Becker, Sarah Burns, Megan Chao, Dan

Cogan, Rocky Collins, Fred Grinstein, Alexis Morrell, Megan Robertson, Matt Wolf, and
Sharon Wood

AI Technologists Sharad Goel (Computational Policy Lab) and Streeter Phillips (Tech Investor)

Members of the Alliance of Documentary Editors (ADE) Derek Boonstra, Cherree Dillon, Kyle
Crichton, and Davide Fiore

Membership of the Archival Producers Alliance

From June to September 2024, APA members solicited endorsements from filmmakers, production
companies, funders, and distributors.
Dr. Patricia Aufderheide (Founder and Senior Research Fellow, Center for Media & Social Impact,
American University) came on as Senior Advisor of the organization, and the APA assembled an
Advisory Board for the GenAI Initiative. In addition to Dr. Aufderheide, that Board includes:

Megan Chao, Documentary Producer & Editor
Katerina Cizek, Documentary Filmmaker & Artistic Director, Co-Creation Studio, MIT
Jon-Sesrie Goff, Filmmaker and Program Officer, Ford Foundation
Reuben Hamlyn & Sophie Compton, Filmmakers and Co-Directors of Another Body
Robert Mahari, JD-PhD, Harvard Law School & MIT Media Lab
Chris Perez, Partner, Donaldson Callif Perez, LLP
Talia Rosen, Senior Director of Standards & Practices, PBS
Loralee Sundra, Associate Director, UCLA Doc Film Legal Clinic
Shana Swanson, Supervising Producer, ITVS

The Camden International Film Festival / Points North Institute invited the APA to present the Best
Practices document at their fall festival, and a travel grant from the LEF Foundation provided APA
Co-Directors funds to do so in person. A generous launch grant from the The Jonathan Logan
Family Foundation will provide support for the continued work of this initiative.

Endorsing Organizations as of September, 2024