Why This Matters Now

Generative AI is quickly moving from experimental technology to an everyday tool used in workflows, including concept art, previsualization, subtitling, and marketing. Meanwhile, regulations are taking shape in Hollywood and beyond. The Writers Guild of America (WGA) has established rules in the 2023 contract to ensure that writers are credited as authors and that AI does not replace human writing in a way that could affect compensation (WGA MBA 2023). SAG-AFTRA has also negotiated specific guidelines for digital replicas and synthetic performances, primarily focused on consent and compensation (SAG-AFTRA TV/Theatrical). At the same time, regulators are enhancing transparency and safety, especially with the EU AI Act (EU AI Act explainer).

Given this backdrop, Netflix’s guidance provides filmmakers with a practical framework for the appropriate use of generative AI in productions commissioned by the platform, emphasizing human oversight, consent, and legal compliance. This report was first highlighted by IBC (IBC News).

What Netflix’s Guidance Covers

While the specific language targets Netflix-commissioned productions and may evolve, the core themes echo broader industry and regulatory expectations. Here are the main pillars to keep in mind:

• Human Authorship and Accountability: AI can assist, but humans are still the creative authors and decision-makers. All AI-assisted work involved in a production must be evaluated by qualified individuals for accuracy, quality, and safety. This supports the WGA’s position that AI cannot displace human writers as credited authors (WGA MBA 2023) and aligns with the U.S. Copyright Office’s guidance that states copyright protects only human authorship (USCO AI guidance).
• Consent, Control, and Compensation: Productions must obtain informed consent and document terms before generating or altering a person’s likeness or voice, in accordance with SAG-AFTRA’s provisions on digital replicas (SAG-AFTRA). Extra caution is necessary for background actors, stunt performers, and voice actors.
• Respect for IP and Training Data: Avoid using tools that are trained on or produce content from unlicensed data. Ensure that vendors can provide documentation of sources and licenses. This minimizes the risk of infringement and adheres to regulatory efforts toward transparency and origin tracing (EU AI Act).
• Data Security and Confidentiality: Confidential materials, such as scripts and casting lists, should not be uploaded to public models. Use approved tools, secure environments, and minimize data sharing. The UK ICO’s guidance on AI and data protection offers useful control measures, including access limits and audit trails (ICO).
• Transparency and Labeling: Clearly indicate when synthetic media is utilized and keep internal records of tools, prompts, and outcomes. The EU AI Act mandates transparency for certain generative uses, requiring labeling of AI-generated or manipulated content in contexts where there are risks involved (EU AI Act).
• Fairness, Bias, and Safety Checks: Assess AI outputs for representational harms, factual inaccuracies, and safety concerns. Document any reviews and corrections, and don’t overly depend on AI outputs without human verification. The U.S. FTC also emphasizes the need for substantiation in AI claims and vigilance against misinformation in AI content (FTC).
• Vendor and Tool Due Diligence: Vet vendors for rights, compliance, and security. Favor tools that support watermarking or provenance signals and provide enterprise-level controls. The Partnership on AI advises adoption of disclosure and metadata practices for synthetic media (PAI).

Practical Examples: What is Typically OK vs. Risky

Commonly Acceptable with Safeguards

• Development: Creating mood boards or concept iterations with fully licensed datasets and enterprise tools, followed by human curation.
• Previs and Shot Planning: Quick previs, animatics, or location visualization, maintaining a clear distinction from final photography choices.
• VFX and Post: Cleanup, rotoscoping, upscaling, and background elements, provided artists retain control and approve the final results.
• Localization: Assisting with subtitling or dubbing where translations and performances are reviewed by qualified humans, using licensed voices with consent for any voice synthesis.
• Quality Control and Accessibility: Automated checks for technical issues or generating descriptive audio drafts, which are then refined by humans.

High-Risk or Likely Prohibited

• Unlicensed Training or Cloning: Using a person’s likeness or voice without explicit consent and a written agreement, or employing models trained on unlicensed copyrighted works.
• Public Models for Confidential Material: Uploading scripts, rough cuts, or unreleased materials to public chatbots or tools that retain prompts or outputs.
• Undisclosed Synthetic Media: Incorporating AI-generated scenes or assets without proper documentation and appropriate disclosure where required by law or platform policy.
• Replacing Credited Authorship: Substituting AI outputs for human-written script pages in ways that are inconsistent with the WGA agreement or local labor contracts (WGA MBA 2023).

How the Guidance Fits with Wider Rules

Netflix’s approach generally complements existing contracts and laws rather than creating new ones:

• Union Agreements: The provisions set by WGA and SAG-AFTRA continue to set the standard for covered productions in the U.S., especially concerning authorship, consent, and compensation (WGA) (SAG-AFTRA).
• Copyright and IP: Productions must secure rights for any training data or reference materials and adhere to national copyright laws. The U.S. Copyright Office has clarified the process for registering works that include AI-generated content, emphasizing human authorship (USCO).
• Transparency and Safety: The EU AI Act introduces obligations for generative systems, including the necessity for transparency about AI-generated or manipulated content in specific contexts, along with added risk management for higher-risk uses (EU AI Act).
• Privacy and Security: Data protection authorities like the UK ICO provide actionable steps for the lawful and secure use of AI, which includes conducting data protection impact assessments and overseeing vendors (ICO).

What Producers Should Do Next

Whether your project is commissioned by Netflix or any other platform, following these practices will help you stay compliant and efficient:

• Develop a straightforward AI use plan that outlines the tools, purposes, and review processes. Maintain an internal log of prompts, training sources, approvals, and changes.
• Obtain consent for any likeness or voice work. Ensure clear contract language and track expirations, scope, and compensation.
• Evaluate vendors for licensing, provenance, security, and data handling. Opt for enterprise solutions that include audit logs and opt-out capabilities.
• Protect confidential materials by utilizing private or on-premises deployments whenever possible, and avoid using public tools for sensitive content.
• Label synthetic media where necessary and maintain provenance metadata throughout the production process.
• Conduct human reviews for accuracy and safety, documenting any corrections made along the way.

Conclusion

Generative AI is quickly becoming a standard tool in filmmaking, but trust hinges on ensuring human control, consent, and transparency. Netflix’s guidance establishes a pragmatic set of ground rules that align with union agreements and evolving regulations. By implementing clear workflows now, you can harness the creative potential of AI while safeguarding your story, your team, and your audience.

FAQs

Does this mean AI can write scripts for Netflix?

No. Human authorship is still crucial. The WGA agreement protects writers’ credits and compensation, and AI use cannot violate these terms (WGA MBA 2023).

Can productions use AI to clone an actor’s voice or likeness?

Only with explicit, informed consent under a written agreement and within the framework of union rules and local laws. Uncompensated or undisclosed cloning is high-risk and likely prohibited (SAG-AFTRA).

Are public AI tools allowed?

Avoid using public tools for confidential materials. Utilize vetted, enterprise solutions that prioritize security, clear licensing, and opt-out provisions, while keeping audit logs (ICO).

Do we need to disclose AI-generated content to viewers?

Disclosure requirements depend on your jurisdiction and the context. The EU AI Act includes transparency obligations for synthetic media in specific situations, and some platforms and festivals enforce their own rules (EU AI Act).

How should we document AI use?

Maintain a detailed AI log that captures tools, versions, purposes, prompts, data sources, approvals, human reviewers, and any output changes. This practice supports compliance, attribution, and quality assurance.

Sources

1. IBC – Netflix Publishes Generative AI Guidance for Filmmakers
2. WGA MBA 2023 – AI Provisions Summary
3. SAG-AFTRA TV/Theatrical – AI and Digital Replicas
4. European Parliament – Artificial Intelligence Act Explainer
5. U.S. Copyright Office – Artificial Intelligence Initiative
6. UK Information Commissioner’s Office – AI and Data Protection
7. U.S. FTC – Keep Your AI Claims in Check
8. Partnership on AI – Responsible Practices for Synthetic Media