Rules for generative AI in the design studio: a governance checklist

KEY TAKEAWAYS Summary by the editors

1. Generative AI governance in a design studio should cover six areas: approved tools, data uploads, intellectual property, disclosure and provenance, skills, and accountability.
2. Under the EU AI Act, AI literacy obligations have applied since 2 February 2025, so organisations deploying AI need to ensure staff using it have sufficient understanding.
3. Article 50 of the EU AI Act applies from 2 August 2026 and requires machine-readable marking of synthetic images by providers and disclosure of deepfakes by deployers.
4. Studios should document human creative input in AI-assisted designs, because the US Copyright Office concluded in January 2025 that prompts alone do not establish authorship.
5. Good rules are short, name an owner for each decision and are reviewed at least once a season as tools and laws change.

A design studio needs a short, enforceable set of rules for generative AI that answers six questions: which tools are approved, what material may be uploaded, how intellectual property is protected, when AI use is disclosed, what training staff receive, and who signs off. Without them, AI use spreads through personal accounts and creates confidentiality, IP and compliance risks the brand cannot see.

Why does a design studio need generative AI rules?

Use is already widespread. McKinsey's State of Fashion research reports that more than 35 percent of fashion executives surveyed already use generative AI, with image creation among the routine tasks. In design teams this often begins informally: a designer tests a consumer image tool for a mood board, uploads an unreleased sketch, or uses a generated print as a starting point. Each step can be harmless, but without rules no one knows which unreleased designs have left the company, which prints have uncertain origins, or which tools' terms allow the provider to reuse uploads.

Regulation adds a further reason. The EU AI Act's general provisions, including AI literacy, have applied since 2 February 2025 according to the Commission's implementation timeline, and Article 50 transparency obligations apply from 2 August 2026.

What should the governance checklist include?

Read also

How is AI used for colour, prints and materials in fashion?

Which tools and data uploads should be allowed?

• Approve tools by account type, not just brand name: enterprise terms often differ from consumer terms on data use.
• Check whether the provider uses uploads or prompts to train models available to others, and prefer settings or contracts that exclude this.
• Classify material: public references, brand archive, unreleased collections, licensed third-party artwork and personal data such as model images.
• Forbid uploading licensed third-party prints or images of identifiable people unless rights and consent clearly cover it.
• Keep a register of approved tools with an owner and a review date.

The tool register should be practical rather than exhaustive. For each approved tool it should state the permitted uses, the account type, whether outputs may be used commercially, whether uploads are retained or used for training, and who in the studio is the contact for questions. Designers should be able to check it in a minute.

How should studios handle IP and authorship?

Two risks run in opposite directions. Outwardly, AI output may resemble someone else's protected design. Inwardly, the brand may struggle to protect AI-assisted work as its own. On the second point, the US Copyright Office concluded in January 2025 that prompts alone do not provide sufficient human control to make the user an author, while perceptible human expression and creative selection, arrangement or modification of AI output can be protected. Rules should therefore require originality checks before sampling and a simple record of what the designer contributed.

When should AI use be disclosed?

Studios should also decide how AI-assisted work is recorded in their own systems. A simple field in the design or approval record, stating whether generative tools were used and which ones, costs little and answers questions later from legal, from retail partners or from customers. It also lets the studio see where AI is genuinely used, which is useful evidence when the rules are reviewed.

Article 50 of the EU AI Act requires providers of systems generating synthetic images to mark outputs in a machine-readable format, and deployers of AI that creates deepfakes to disclose this, with lighter rules for evidently artistic or creative work. The Commission's timeline notes a transition until 2 December 2026 for certain systems already on the market. Studios should keep provenance metadata intact rather than strip it, and agree internally when AI-generated imagery, for example synthetic models in campaign visuals, is disclosed to consumers. Content provenance standards such as C2PA, whose Content Credentials record a file's origin and edits, can support this.

How should a studio roll the rules out?

1. Survey current use honestly, without penalties, to see which tools designers already rely on.
2. Draft the one-page policy and the tool register with design, legal and IT.
3. Run short training sessions so every user understands the rules and the tools' limits.
4. Move work onto approved accounts and close personal-account use for company material.
5. Add an AI field to design and approval records so AI-assisted work is visible.
6. Review the rules each season and after any incident, change of law or major tool update.

Read also

Who owns an AI-generated fashion design? Copyright and IP explained

Who is accountable when AI-assisted designs go wrong?

The brand is. A supplier, regulator or rights holder will not accept that a tool produced an error or a copy. Governance therefore needs named people, not only principles: someone approves tools, someone signs off AI-assisted designs before sampling, and someone owns incidents. McKinsey's research stresses that companies must protect the creativity at the core of fashion while adopting AI; clear rules are what allow designers to experiment without putting that creativity, or the brand's rights, at risk.

Frequently asked questions

Do fashion brands need an AI policy for designers?

Yes, if designers use or may use generative tools. A policy clarifies which tools are approved, what may be uploaded, how IP is checked and who signs off. It also supports the EU AI Act's AI literacy expectations, which have applied since 2 February 2025.

What should be in a generative AI policy for a design team?

At minimum: approved tools and account types, rules on uploading unreleased or third-party material, originality and IP checks, disclosure and provenance rules, required training and named approvers. A short register of tools with owners and review dates keeps it practical.

Does the EU AI Act require labelling AI-generated images?

Article 50 requires providers of systems that generate synthetic images to mark outputs in a machine-readable way, and deployers to disclose deepfakes, from 2 August 2026. Evidently artistic or creative works have lighter disclosure rules. Brands should check how their tools and campaigns are affected.

Can designers use public AI image tools for commercial work?

Only within the studio's rules. Teams should check each tool's terms on commercial use and on whether inputs train shared models, avoid uploading confidential or third-party material, and document human creative input. Enterprise accounts usually offer clearer data terms than personal ones.

GuideThe complete guide to AI in fashion design and product developmentRead the complete guide