USE OF AI

We see AI as a creative collaborator, not a replacement for human craft. Every deliverable is shaped, edited and signed off by humans. We are transparent with clients and audiences about how AI is involved, and we refuse work that uses AI to deceive or to harm.

• Human in the loop: a senior creative reviews every output before it ships.
• Transparency: clients always know which parts of a deliverable were AI-generated.
• Consent: we do not clone real voices, faces or styles without documented permission.
• Provenance: we keep records of the tools, prompts and source assets used in production.

We work with a curated stack of commercial AI tools across language, image, video, audio and code. Our default position is to use enterprise-grade plans with data-handling controls and no training on client inputs.

Where we use open-source models, we run them in environments we control. We avoid models with unclear licensing, training data of disputed origin or known safety issues.

Client briefs, assets and feedback are treated as confidential. We do not submit them to consumer AI products. We only use providers that contractually agree not to train their models on our inputs, and we limit access to project data on a need-to-know basis.

We generate assets in a way intended to be commercially usable in your market. Where AI outputs may have limited copyright protection in certain jurisdictions, we say so up front and propose practical mitigations — for example combining generated work with original human elements, or using licensed source material.

We do not generate the face, voice or recognisable style of a real person without prior written consent from that person or their authorised representative. Consent records are retained for the life of the work.

We disclose AI involvement to clients in our proposals and statements of work, and we recommend appropriate audience-facing disclosures (for example labelling synthetic media on social posts) when relevant rules or platform policies apply.

We will not produce content that is deceptive, defamatory, sexually explicit involving minors, that promotes self-harm or that targets individuals or protected groups with abuse. We may decline projects, in whole or in part, when these lines would be crossed.

AI capabilities, regulations and norms are changing quickly. We review this policy at least quarterly and update it whenever our tooling or practice changes meaningfully. Material changes are noted at the top of this page.