Our Generative AI Policy

Deep Fusion Films: Our Approach to Generative AI

Leading with Integrity in a Changing Industry

At Deep Fusion Films (DFF), we believe the future of storytelling lies at the intersection of innovation and responsibility. Generative AI (Gen AI) is already transforming how we create, deliver, and experience content. It offers powerful new possibilities - but also brings with it serious ethical, creative, and environmental questions.

We don’t shy away from those questions. Over the past 18 months, we have been a visible and vocal advocate for ethical AI in the creative industries - speaking at WIPO events, giving evidence to a UK Government Select Committee, and helping shape policy for industry unions and representative bodies. We are not passive adopters of AI. We are actively helping define how it should be used - and how it should not.

This document sets out how we use Gen AI at Deep Fusion Films. It is a living commitment: not a marketing statement, but a framework of accountability to our industry, our collaborators, and our audiences.

1. The Reality of the Industry Today

Broadcasters and networks are already leaning into Gen AI - especially when it:

• Speeds up valid processes

• Reduces cost or post-production time

• Is used transparently and doesn’t mislead audiences

To stay relevant, trusted, and commissionable, we must evolve with the industry - but not at the expense of ethics or trust. That’s why we’re choosing to lead from within. We use Gen AI responsibly, with human direction, visible boundaries, and creative intent at the core. We acknowledge the imperfections within the system and are actively involved in building a future with equitable remuneration for creatives, data provenance and copyright protections.

2. Why We Use Gen AI

We use Gen AI to enhance and accelerate pre-existing production processes - not to replace the work of human creatives.

Our current uses include:

• Animating licensed stills

• Converting artwork or illustrations into photorealistic images

• Generating background plates where traditional filming is impractical

• Assisting with transcription, subtitling, and localisation

• Speeding up previsualisation and asset development

These are roles that follow a well-established lineage of production tools - like colourisation, motion interpolation, rotoscoping, and digital VFX. Gen AI is the next step in that evolution.

We do not use Gen AI to generate entire films, stories, or performances from scratch unless it is for Research and Development. Our internal development work may be released online for the purpose of informed audience feedback.

3. Human Creativity Is Always Central

Every Gen AI output at DFF is:

• Initiated by human creatives

• Directed, approved, and integrated within a human editorial vision

• Reviewed for tone, accuracy, and audience perception
  

We never use Gen AI to:

• Replace scriptwriters, editors, or performers

• Mimic living individuals without explicit consent

• Generate misleading visual material for factual productions

When AI is used in the creative process, we credit the human, not the tool. AI assists. It never authors.

4. Consent, Estates, and Likeness Rights

We do not and will not use Gen AI to recreate the likeness, voice, creative style, or identity of any living person without their explicit, contractual consent.

Where deceased individuals or estates are involved, we obtain written permissions and engage in equitable partnership models to ensure their legacy is respected.

5. Addressing the Training Data Dilemma

We acknowledge that many Gen AI tools have been trained on vast datasets that likely include copyrighted content or third-party materials without consent. This is a fundamental and unresolved issue - one that requires transparency and reform.

At the same time, Gen AI models must learn from real-world patterns to function. An AI doesn’t understand how trees move, how faces light, or how emotions play out unless it has seen millions of examples. That’s modelling, not mimicry.

There’s nothing copyrightable about gravity or motion - but we recognise that creative work is part of what taught these systems how to see.

That belief directly informs our reinvestment strategy (6).

Additionally, while Gen AI is capable of imitation and replication, and therefore capable of infringing on existing copyright or IP, we will never knowingly use Gen AI tools or processes to create new material based on existing intellectual property without permission. 

We carry out due diligence to clear the rights in content we use in AI assisted processes

6. Reinvesting in the Industry

We believe that if AI tools benefit from the creative ecosystem, they should help sustain it.

That’s why we:

• Donate a portion of profits from AI-assisted productions to the TV & Film Charity, supporting industry professionals in need

• Deliver a dedicated training programme in partnership with the National Film and Television School (NFTS), focused on the responsible, creative, and practical use of Gen AI in production

Through both financial support and knowledge sharing, we aim to ensure the future of the industry is shaped by those who understand it - not just those who code it.

7. Transparency and Guardrails

We operate with full transparency. On every project, we:

• Declare the use of Gen AI to commissioners, collaborators, and clients from the outset

• Track all use of Gen AI internally via a tiered risk framework

• Label and disclose AI-generated or enhanced materials clearly in documentation or credits

• Disclose use to the audience where it may impact understanding - especially in factual or documentary content.

Our tiered framework:

• Tier 1 – Technical enhancement only (e.g. transcription, stabilisation)

• Tier 2 – Augmenting licensed material (e.g. animating stills, background creation)

• Tier 3 – High editorial sensitivity (e.g. recreating people, historical events) - used only with sign-off and clear labelling

We have evaluated the Gen AI tools we use and we never commercially use software or platforms where the materials we input - such as scripts, stills, or footage - are used to train or improve that provider’s models. We work only with services that offer clear contractual assurances that user-uploaded content remains private and proprietary.

We are also developing internal tools to ensure this tracking is consistent, reviewable, and auditable.

8. Responsible Use in Factual and Historical Storytelling

We do not use Gen AI to generate misleading “archive,” or create visuals that could distort real events. 

If synthetic material is used in factual work:

• It will be clearly labelled

• It must be justifiable

• It must not risk misleading the viewer

History is not a prompt. We treat it with respect.

9. Our Role in the Labour Market

We do not use AI to cut corners or reduce headcount.

Gen AI should support, not replace, skilled labour. We build our workflows to keep editors, writers, designers, and performers in the loop - and to free them from repetitive or non-creative tasks.

We will not use AI to eliminate core creative roles purely for cost savings.

10. Environmental Awareness

AI systems - especially visual generation - are compute-intensive. We are committed to reducing the environmental cost of our work.

We:

• Prioritise efficient models and workflows

• Minimise unnecessary rendering or compute cycles

• Work with partners to explore energy offsetting

In addition, we have:

• Created an internal environmental impact calculator to measure our AI energy use on a per-project basis

• Commissioned an internal study into the full environmental cost of our AI pipeline, with the intention to publish and share insights with the wider industry

Innovation must also be sustainable. We’re making sure it is.

11. A Living Commitment

This policy is not a statement of perfection - it’s a declaration of direction.

We continually take advice from respected third-party professionals - including legal, insurance, and policy experts - to inform and test our thinking. This ensures we are not operating in a bubble or echo chamber. We welcome challenge, scrutiny, and new perspectives, and we actively learn from them. Our position continues to evolve because we believe ethical leadership requires constant listening, not static rules.

As technology evolves, so will we. We will continue to:

• Refine our internal tooling and tracking

• Build AI products that embed ethical principles by design

• Participate in public discourse, industry panels, and policy conversations that shape the future of creativity

We are not just responding to the wave - we are helping navigate it.

The Bottom Line

We will use Gen AI.

We will use it responsibly.

We will use it to speed up work - not to replace people.

We will give back to the creative ecosystem.

We will build tools with accountability in mind.

We will educate others and challenge ourselves.

We don’t hide what we do - we label it, explain it, and stand behind it.

AI won’t define Deep Fusion Films.

But the way we use it will.