Ethical AI

AI is an execution layer, not a decision-maker of record. Every campaign that reaches the market carries a named human approver. We do not build features whose purpose is to remove that person from the loop.

Assets produced by the platform are identified as AI-generated in the internal asset record, including the agent, the model family and the sources retrieved. Clients receive this provenance with every deliverable and may disclose it publicly as their market or regulator requires.

Agents must not fabricate product claims, performance figures, endorsements, awards or testimonials. Any factual claim in a generated asset must trace to a client-supplied source. Claims that cannot be traced are flagged for human substantiation before approval.

• 01Audience segmentation must not target or exclude people on the basis of protected characteristics where that use is unlawful or discriminatory.
• 02Generated creative is reviewed for stereotyping and for representation that is degrading or demeaning.
• 03Where a market imposes stricter rules on targeting or representation, the stricter rule applies.

Agents operate on the brand and campaign data required for the task. We do not build behavioural profiles of individual consumers, and we do not enrich client data with third-party personal data for targeting purposes.

Content directed at children, and advertising in regulated categories such as financial services, health, alcohol and gambling, is subject to mandatory human review with category-specific checks before it can be exported.

Agents work from client-supplied brand assets and licensed inputs. The platform does not knowingly reproduce third-party trademarks, copyrighted characters or protected artistic styles, and outputs that risk doing so are blocked at the brand safety layer.

If a generated asset causes harm or breaches these commitments, clients may request its removal from the platform, an explanation of how it was produced, and a record of the approval chain. Contact [email protected].