Generative AI In Media Production: Legal Considerations For Canadian Film And TV Producers

Commissioned by the Canadian Media Producers Association (CMPA)

General Introduction

The last two years have seen an unprecedented explosion of online platforms providing easily accessible generative artificial intelligence tools that create content at a scale, quality and cost that can be significantly disruptive innovations for media production. The CMPA recognizes that while generative artificial intelligence ("generative AI") has the potential to provide benefits to producers, these tools also present legal, moral, and practical considerations. Can I own copyright in this material? What are the data privacy concerns? What are the social considerations? What are the union and guild requirements?

This document is intended for a broad audience of producers and their business and legal affairs departments. It is not intended as legal advice or prescriptive guidance on the use of AI, but rather as a set of considerations and risk mitigation strategies meant to identify areas of particular sensitivity and concern for media producers. Finally, because this landscape is evolving rapidly, this document should be considered a signpost on the road, not a definitive destination.

Overview of contents

This discussion uses the generic term "AI" to refer to generative AI services and platforms that can be used to create new content, including text, images and sounds, in response to prompts provided by human users. In the media production industry, AI has a wide range of potential applications, from creating or modifying screenplays to generating art used for set dressing, to producing entire musical soundtracks to accompany filmed (or AI-generated) images, to creating on-screen performances of non-existent "actors" (which may or may not deliberately echo the look or style of actual living actors), and more besides.

This guidance does not address all forms and uses of AI – rather it is focused on generative AI, and specifically the "outputs" of generative AI: content that can be incorporated into or serve as a media production and which can be either seen or heard by a viewer in the same way that they could see a drawing or hear a song created by a human artist.

[This discussion does not delve into the technical aspects of AI and related technologies such as machine learning and large language models, and is not focused on technical or "embedded" uses of artificial intelligence in software or services which are largely invisible to an end user (such as data analytics). Instead, it concerns itself with the use of AI to generate outputs that can themselves be manipulated and included in the media production workflow.]

The use of AI platforms presents both opportunities and risks. Canadian producers must learn how to effectively and safely use AI if they are going to remain relevant and competitive in a globalized industry. This guidance aims to facilitate analysis and risk assessments by media producers, with the goals of (1) identifying potential constraints on the use of AI-generated content and (2) setting out practices and steps to be taken when interfacing with AI platforms and services.

This guidance is presented in the form of a series of questions or checklists that can be reviewed when considering the use of AI platforms and services to create AI-generated content in connection with media production. It is not intended to provide definitive answers (since those are rare when it comes to AI) but instead is intended to prompt critical assessments of the use of AI in a way that is sensitive to the particular context of the media production industries. This guidance does not offer comments on the suitability of any particular AI platform or service; the terms of use and licensing agreements used by those platforms and services are revised frequently, the legal environment remains fluid, and collective bargaining agreements are being renegotiated, and so each AI platform and services needs to be reviewed at the time the assessment of the service or platform is being made. Further, there are many different types of AI services and platforms, ranging from free-to-use and publicly available to bespoke products designed and tailored specifically for use in connection with media production activities. The framing contained in this guidance is necessarily generalized, so users will need to assess any particular AI platform or service on its own terms.

We offer the following additional suggestions:

Stay current. The pace of developments in the AI space means that this guidance may be outdated when you use it. If you are using AI, consider tasking one or more individuals in your organization with keeping abreast of recent developments and, just as importantly, how other participants in the media production ecosystem are handling AI.

Be transparent. Be as transparent as possible with third parties about the use of AI in connection with the development, production and post-production of a project. The use of AI can impact the rights of those third parties and your obligations to those third parties. Relatedly, audiences will be interested in knowing when and how AI has been used in connection with a project, and its use may impact critical and audience reception, so consider disclosing the nature and extent of such use.

Be proactive. Develop an internal policy stipulating whether, when, and how AI is to be used in the development, production and post-production processes. Regularly review and revise the policy to ensure it remains responsive to developments and up-to-date. Ensure that due diligence is completed on all internal uses of AI prior to their use.

Be responsible. AI does not and should not be viewed as sweeping away relevant legal and ethical obligations and concerns. Approach AI taking into account considerations about matters such as intellectual property rights and data protection and with a view to productively using AI to enhance human creativity, instead of being a substitute for it.

Gating questions – Are we allowed to use AI?

E&O Insurance: Does the E&O insurance policy for the project expressly permit or exclude content created using AI?

• If the policy permits the use of content created using AI, is there a requirement to notify the insurer in order to ensure coverage or other conditions with respect to such use of AI in order for the coverage to apply?
• If the policy is silent, do not assume that the policy will cover the use of AI.
• Consider requesting confirmation from the insurer regarding the use of AI – note that there are risks inherent both with remaining silent about the use of AI (such as a later denial of coverage) and notifying an insurer about using AI (such as a refusal to issue coverage in the first instance).

Distribution / Licensing agreements:

Do the distribution agreements, broadcast licence agreements and other exploitation agreements applicable to the project permit or exclude the use of AI?

• Flag: Note that prohibitions can take different forms: there may be an explicit prohibition against incorporating AI-generated content into production deliverables or the agreement may contain representations and warranties that are inconsistent with the use of AI-generated content (e.g., a requirement that all deliverables be "original" and capable of being protected by copyright – something that AI-generated content does not satisfy).

Tax credits / Funding / Incentives: Are grants, tax credits or other incentives being claimed for the project? Do the eligibility rules for the incentives explicitly exclude or permit the use of AI-generated materials in the project?

• Various funding bodies have published AI policies or have modified their policies to address AI, including the Canada Media Fund, Ontario Creates and Telefilm Canada – those policies will almost certainly be updated from time to time, so remember to check them at the time of application.

Financing requirements: If funds are being borrowed to produce the project, do the terms of the loan impose any restrictions on the use of AI or require any notification to the lender of such use?

• Determine whether there are any relevant representations and warranties that pertain to copyright (e.g., in a copyright mortgage or security agreement).

Disclosure of confidential or sensitive information: Does the information (text, image, data) that will be used to prompt the AI contain any confidential information that should not be disclosed for strategic reasons (e.g., character designs, plotlines, etc.) or personal information that should not be disclosed for privacy compliance reasons (e.g., personally identifying information about an individual)?

• Confidentiality can be critical during development of a project, so bear in mind the risk of development content being "leaked" or otherwise becoming publicly available through the use of AI platforms and services. Some AI platforms and services have terms of use that give the platform/service provider the unrestricted right to use any material used to prompt the AI – meaning that text prompts and visual prompts can not only be used to "train" the AI but can also possibly end up as "outputs" of the AI, with the results that your development material not only becomes publicly known but is subject to a significant ambiguity as to what a user of the AI platform/service can do with it. In short, you may end up in a situation where your "secret" character design has inadvertently been made publicly available for anyone to use without permission or payment

Production counsel: Has the use of AI been approved by production counsel for the project?

Clearance questions – When we use AI, what should we consider?

Terms of use / Licence terms: The terms of use and/or user licence agreement for the AI platform or service must be reviewed for suitability for use with audio-visual productions.

• This review should, in general terms, be comparable to the review undertaken for platforms that provide "stock" materials (such as images, film clips and music).
• If the review indicates that the terms are not suitable for use in audio-visual productions (e.g., the terms restrict use of the AI-generated output to specific territories, instead of "worldwide") and there remains a need/desire to use the AI platform or service, determine whether it is possible to obtain from them a bespoke arrangement (similar to how some stock photo providers, such as Getty, will provide a "Film & TV Rider" that supplements their standard terms and conditions).
• Assess whether the AI provider has made a commitment to ensure that its outputs do not contain or demonstrate any impermissible bias (e.g., you want to avoid AI platforms that, when prompted to generate images of "doctors" only generate images of older white males).
• Assess whether the AI provider has appropriate safeguards in place to protect data privacy with respect to prompts and other information collected by the provider
• Among the terms to consider in reviewing the terms of use and/or user licence agreement of an AI provider:
  • Does the platform or service provide an indemnity if an infringement claim arises as a result of using content generated by the platform?
  • Does the platform or service obtain any rights to use the prompts (particularly image prompts) provided by the user to the platform or service?
    • If the platform/service does obtain such rights, be aware that this can pose significant risks to a production, such as the production no longer having exclusive rights or control to the content used to prompt the AI – meaning that other people can use your prompts without having to get your permission or make payments to you.
  • Does the platform or service relinquish all rights in the content generated by the platform or service and/or grant all rights (if any) in the content to the user?
    • If the platform/service retains rights in the AI-generated content, that means you may need to get their permission or make certain payments to them if you exploit the content generated by their AI.
  • Are there any restrictions on the use of content generated by the platform or service?
    • For example, restrictions on use in connection with certain activities (such as "adult" content) or on the number of copies that can be made or the media, territory or time period in which content can be used – such restrictions can significantly hamper your ability to exploit your production, and can constitute a breach of contract under distribution and licensing agreements.
  • Are there any representations or warranties regarding the ownership of the content and/ or the power to grant the rights granted by the platform or service in the terms of use or licence agreement?
    • Beware of AI platforms and services that make no representations or warranties and offer no indemnities regarding the content they generate – if something goes wrong with the content (e.g., the content is found to infringe someone else's copyright), you will have no recourse against the provider of the platform/service.
  • Are there any on-screen credit obligations triggered by use?
    • A production may need to specifically identify the provider of the platform/ service in the end credit roll in order to be in compliance with the terms of use.
  • Is there a waiver of moral rights, if any, in the content?
    • Many distributors and licensees of media productions will require waivers of moral rights for content included in a production – so the absence of a waiver of moral rights can lead to delivery problems.
  • Is there a waiver of injunctive relief by the platform or service?
    • Many distributors and licensees of media productions will require waivers of injunctive relief from content providers – so the absence of a waiver of injunctive relief can lead to delivery problems.
  • Is any licence granted or rights transfer revocable for any reason?
    • A revocable licence means that the AI-generated content would have to be removed from your production (requiring edits and re-cutting) – potentially even years after the production has already been delivered.

Guild / Union constraints: Is the production subject to any guild or union collective agreement that contains provisions relating to AI? At the time of writing, The Writers Guild of Canada (WGC), The Writers Guild of America (WGA), The Screen Actors Guild-American Federation of Television and Radio Artists (SAGAFTRA) and the IATSE collective agreements impose restrictions on the use of AI, as follows:

• WGC
  • Producer must disclose to the writer if any materials given to the writer have been generated by AI. In such an event, the AIgenerated material shall be deemed to be "Source Material."
  • Effectively, such AI-generated Source Material shall be recognized and treated no differently than if the Source Material were created by a human (e.g., a novel with a traditional author), as the writer will be entitled to all applicable credit and financial entitlements under the WGC IPA.
  • AI-generated material shall not be considered "Development Materials" or "Script Material."
  • Producer may not use Development Materials / Script Material for which writer retains copyright to train AI unless Producer does the following:
    • Clearly set out that they are using such material for training purposes;
    • Provide a description of the intended use; and
    • Obtains conspicuous consent (i.e., including a separate space for writer's acknowledgment to the explicit provision)
  • The WGC / CMPA acknowledged that this is a nascent issue (specifically with respect to training AI) and the parties agreed to meet at least semi-annually during the term of the IPA to discuss developments.
    • [Note: As of August 2024, the WGC is the only Canadian guild with any relevant collective agreement provisions, but that may change imminently.]
• WGA
  • AI cannot be considered a writer/author for productions.
  • Producer must disclose to the writer if any materials given to the writer have been generated by AI or incorporate AI-generated material, provided that the AI-generated material cannot be Source Material (meaning that a writer's "Separated Rights" or credit entitlements under the WGA cannot be undermined).
    • In contrast to the treatment under the WGC IPA, AI-generated material on which a script is based is treated differently than human-generated Source Material (and, effectively, AIgenerated material is ignored for the purposes of determining a writer's separated rights).
  • Writers cannot be forced to use AI by a producer.
  • Writers can choose to use AI if the producer consents, provided that the writer follows applicable company policies.
  • WGA reserves the rights to prohibit the use of writers' material to train AI (i.e., such use is permitted at the moment, but the WGA may prohibit it later).
• SAG-AFTRA
  • Producer must obtain "clear and conspicuous consent" in order to create a "digital replica" and to use the digital replica unless the audio/video remains substantially as originally scripted, performed or recorded.
  • Producer must pay the performer for the number of days they would have been required to perform scenes using the digital replica.
  • Moving "background actor" digital replica to "principal actor" role requires compensation.
  • Residuals payable for use of digital replica on-screen.
  • Schedule F players are exempt from minimum compensation (entirely negotiable).
  • Separate consent required to use digital replica in a different project.
  • Consent required before significantly digitally altering a performance (excluding standard post-production adjustments and matching lips to a dubbed foreign language).
  • Producers must obtain union consent for use of "synthetic performers" (i.e., nonreplica digital performers)(does not apply to non-human characters).
• The Memorandum of Agreement on the Basic Agreement between IATSE and AMPTP, dated as of August 1 2024, also includes AI guidance, covering the use of new technologies by both producers and employees, work training programs, and ongoing obligations to meet at least semi-annually to discuss the use of AI systems. Producers should consult this MOA if they plan to engage IATSE members.

Contributor / Actor constraints

• For actors and writers, consider any guild/unionlevel constraints (see above).
• Are there any AI-related restrictions contained in the agreement with the contributor/ actor?
  • Example: Restrictions on using the contributions for AI "training" purposes.
  • Example: Restrictions on using AI to modify the contributions.
• Consider contracting to allow certain AI uses (such as "incidental" AI for image enhancing or smoothing, dubbing, etc.).

Concerns re lack of ownership of AI content

Consider whether the production company is comfortable with the prospect of "new" content generated by an AI platform or service being, for copyright law purposes, un-authored, and hence unowned, and hence not protected by copyright – in other words, would the production company be aggrieved if a third party themselves took and used the "new" content and the production company had no recourse against them?

• For example, if an AI platform is used to create the design of a new "hero" character in an animated series, the unrestricted appropriation and use of that character by a third party may pose significant business and financial risks.
• Factual information: If using the AI service to generate a factual statement that will be used in a production, has the accuracy of the statement been confirmed by referencing other sources? AI can generate "hallucinations" that contain incorrect, wholly fabricated or misleading information.
• Reviewing AI outputs: When reviewing the content created by AI, consider the following:
  • Does the AI-generated content infringe copyright (i.e., does the output copy a substantial part of an existing copyright-protected work)?
  • Does the AI-generated content infringe trademark rights (i.e., does the output duplicate or confusingly resemble an existing logo or brand)?
  • Does the AI-generated content infringe any personality-based rights (i.e., does the output copy or resemble the image, voice or other identifying characteristics of a celebrity)?
  • Is the AI-generated content defamatory (i.e., does the output contain a statement about a living individual or a business/ organization that could damage their reputation or portray a living individual as saying something that could damage their reputation)?
  • Is the AI-generated content misinformation (i.e., does the output contain a statement of purported fact that is incorrect or misleading)?
  • Does the AI-generated content contain or exhibit any impermissible bias (e.g., generating an image containing only white males when asked to portray "professionals")?

Contracting practices

• Consider including in your contracts (both with employees and independent contractors) restrictions on the use of AI-generated materials, accompanied by an indemnity for a breach of such restrictions.
  • e.g., prohibition on using AI in the course of rendering services unless specifically approved by the producer.
  • e.g., representations and warranties that all materials submitted under the contract will not include AI-generated content unless otherwise approved in writing.
• If asked to sign a contract that contains restrictions on your use of AI, consider modifying the language to permit uses of AI that should be considered unobjectionable, such as the use of AI to enhance audio or picture quality (e.g., audio pitch correction, video colour correction) or otherwise in compliance with any applicable union/guild restrictions.

Copyright registration concerns

• If registering copyright in the project, determine whether the relevant copyright registry imposes any restrictions on or has provided any formal guidance about the registration of materials created using AI-generated content.
  • The United States Copyright Office has published guidance on registering works that contain AI-generated materials.

Ethical considerations

• Is your use of AI consistent with fostering and enhancing human creativity, instead of replacing it?
• Are you taking steps to identify and remove biased, discriminatory, incorrect or otherwise harmful outputs from your AIgenerated content?
• Is your use of AI explainable and transparent? Can your producing and distributing partners and audiences easily identify whether and to what extent, you have used AI in producing your content?
• Is your use of AI, including the content of your prompts, consistent with protecting the privacy of individuals and their personal data?
• Do you have a process (internal and external) for receiving feedback about your use of AI, assessing that feedback and making appropriate changes in response?
• Are you able to assess the environmental impact of your use of AI, including the carbon footprint of those providers of AI platforms and services you are using?

Assessment checklist

Perform due diligence on the vendor of the AI service or platform and the product being offered:

Review the terms of services / terms of use:

Communications protocols:

Use protocols:

About Dentons

Dentons is the world's first polycentric global law firm. A top 20 firm on the Acritas 2015 Global Elite Brand Index, the Firm is committed to challenging the status quo in delivering consistent and uncompromising quality and value in new and inventive ways. Driven to provide clients a competitive edge, and connected to the communities where its clients want to do business, Dentons knows that understanding local cultures is crucial to successfully completing a deal, resolving a dispute or solving a business challenge. Now the world's largest law firm, Dentons' global team builds agile, tailored solutions to meet the local, national and global needs of private and public clients of any size in more than 125 locations serving 50-plus countries. www.dentons.com

The content of this article is intended to provide a general guide to the subject matter. Specialist advice should be sought about your specific circumstances. Specific Questions relating to this article should be addressed directly to the author.