top of page

Writing a Brand AI Imagery Policy: Rules for Teams, Agencies and Studios

2 days ago
7 min read

Updated: 24 hours ago

A brand policy for AI-generated imagery is a short internal rulebook that tells in-house teams, agencies and studios what they may produce with AI, what they may not, who approves it, how it is labelled and what records must exist. A workable policy covers seven things: permitted uses, prohibited uses, approval stages, disclosure, provenance records, rules for people and likenesses, and an owner who keeps it current. It binds partners through the contract, not through goodwill, so it should be written as requirements a studio can confirm in writing and an agency can pass down. Keep it to a few pages. A policy nobody reads is not a control, and a policy that forbids everything gets ignored. Start with the decisions your teams already face every week, and write the rule that settles each one. This is general information, not legal advice.


Why a policy, and why now


Most brands first meet AI imagery as a series of one-off questions. A social team wants to extend a background. An agency proposes a synthetic model for a regional variant. A studio delivers a campaign and the brand asks, late, how the images were made. Each question gets an answer, but the answers differ by team and by week.


A policy replaces those answers with one position. It also gives agencies and studios something concrete to sign: instead of a general promise to "act responsibly", they confirm specific rules, and the brand can check them.


There is a regulatory reason as well. Since 2 August 2026, Article 50 of the EU AI Act requires deployers of AI systems that create deep fakes to disclose the artificial origin of the content. The Commission's guidelines on transparency obligations make clear that a deployer cannot rely on the machine-readable marking the provider embeds. A brand that publishes photorealistic AI imagery in the EU therefore needs a rule for who discloses, how and where. The details are in EU AI Act Labelling Rules for AI Ad Imagery.


What this policy is not


Two neighbouring documents are easy to confuse with it.


  • A studio vetting checklist is the brand's questionnaire to a partner before hiring. It is covered in How to Vet an AI Production Studio for Brand Safety. The policy is the standing rulebook that applies after the partner is chosen.

  • A legal overview explains which laws touch AI campaigns. The policy is the brand's own decision on how it will behave inside those laws, and often inside a stricter line than the law requires.


The seven parts of a workable policy


1. Permitted uses


State what AI may be used for, in terms of work the teams recognise. One way brands divide it:


  • Open by default: early concepts, mood boards, internal animatics, background extension, retouching support.

  • Open with approval: finished campaign imagery, variants of an approved master asset, product visuals.

  • Case by case: anything involving a person's face or body, a heritage archive, or a product claim.


The point of the tiers is speed. People should see at a glance whether a request needs a conversation.


2. Prohibited uses


Name what is off the table, and give the reason in a clause so the rule can be applied to cases it did not foresee. Typical entries:


  • Showing a product as it is not: a colour, finish, size or function the real item does not have.

  • Imitating an identifiable living person, or a named artist's style, without a written agreement.

  • Images of real events presented as documentary.

  • Synthetic people in contexts where the audience expects a real customer, such as testimonials.

  • Use of any tool or system that has not passed the brand's review.


Keep the list short. Ten clear prohibitions work better than forty overlapping ones.


3. Approval


Say who approves what, at which stage, and what counts as sign-off. A common structure is three checkpoints: concept, work in progress, final. The detail of rounds, version names and sign-off records is described in Review and Approval Workflow for AI Campaign Assets. The policy only needs to fix the principle: a named person at the brand, not the producer, releases the final image, and a human has checked it for hands, text, product accuracy and likeness issues.


4. Disclosure


Set one rule for labelling and say who applies it. Decide at least:


  • When an image is labelled to the public, and in what wording.

  • Who places the label: brand, agency or studio.

  • What the partner must tell the brand, and when. Telling the Client describes that handover from agency to brand.


Write the wording once and reuse it. Ad hoc labels drift, and drift reads as evasion.


5. Provenance and records


Each final asset should have a record that answers four questions: which partner produced it, which AI systems were used, which inputs and licensed references went in, and who approved it. Ask partners to keep the record with the asset and deliver it with the files. What a studio hands over at the end is set out in What You Receive From an AI Production Studio.


Content credentials and embedded metadata can support the record, but they do not replace it. Metadata is easily lost when a file is exported or re-saved, so the paper trail has to exist separately.


6. People and likenesses


This is where brands most often rely on assumption. Rules to consider:



7. Ownership, review and change


A policy needs an owner, usually a role in brand or legal, with a named deputy. Set a review date, because tools and rules move. Say how exceptions are requested and who may grant them, and keep a log of them. Add a short change history so partners can see which version they confirmed.


Making agencies and studios bind to it


A policy that sits on an intranet does not reach the people producing the work. Three habits make it effective.


  • Attach it to the contract. Reference the current version in the brief or master agreement, and require written confirmation of each new version.

  • Pass it down. Where an agency uses a studio, the agency confirms that the same rules run through the subcontract. The structure of that relationship is explained in How AI Production Studios Work With Ad Agencies.

  • Ask for exceptions in writing. A studio that wants to depart from a rule asks first and gets an answer on file.


Agencies and studios see the practical failure points daily, so they are well placed to help draft these rules. Involving them early produces a policy that can be followed.


A short test before you publish it


Hand the draft to someone who was not involved and ask three questions: can they tell whether a given request is allowed, who approves it, and what must be recorded? If any answer takes more than a minute to find, the policy needs editing, not more clauses.


Frequently Asked Questions


What should a brand AI imagery policy include?

It should include permitted uses, prohibited uses, approval stages, a disclosure rule, provenance records for each final asset, rules for people and likenesses, and a named owner with a review date. Seven parts, a few pages, written as requirements that partners can confirm.


Who is responsible for AI disclosure, the brand, the agency or the studio?

The contract decides, and the policy should name one party. Under the EU AI Act, the duty to disclose a deep fake falls on the deployer of the AI system, and who counts as the deployer in a brand, agency and studio chain depends on who uses the system under its own authority. Settle it in writing before production starts.


Does a policy bind our agency and studio automatically?

No. A policy binds partners only when it is part of the contract or confirmed in writing for each version. Where an agency subcontracts to a studio, the agency should confirm that the same rules apply down the chain.


Should the policy ban AI-generated people?

That is a brand decision. Many brands allow synthetic figures for concepts and product views but expect real casting where a customer or a named talent is implied. Whatever you choose, state the rule and the exceptions process, and keep any real person's likeness under a written agreement.


How often should the policy be reviewed?

Set a fixed review date, at least once a year, and review sooner when a law or a major tool changes. Keep a version history so partners can see which version they confirmed.


Is a policy the same as vetting a studio?

No. Vetting is the check before you hire a partner. The policy is the rulebook that applies to every project afterwards, with the studio and the agency confirming it.


Planning a campaign with AI imagery or video? Request an AI campaign.


About The Bridges AI

The Bridges AI is a hybrid AI creative agency: a creative studio and production partner for fashion, beauty, luxury and hospitality brands and for the agencies that work with them. It combines booked real talent, real products and human creative direction with AI production, and offers licensed digital twins of real models, only of models under contract. The Bridges AI works in strategic partnership with Most Wanted Models®, the German model agency founded in 1997, which brings nearly 30 years of experience in bookings, buyouts and usage rights.


Written by the editorial team of The Bridges AI. Last updated November 2026.


Comments


bottom of page