When should you tell the public that AI was involved?
Transparency should not be the final question in your AI process. It should shape the process from the beginning.
In today’s governance note, I want to consider a question that every organisation using AI for public-facing products, services or communications should be asking: when do you tell the public that AI was involved?
This question often arises near the end of the process. The product is ready, the communication has been drafted, and someone asks whether an AI label or disclosure is necessary.
I would bring that question forward.
Deciding how and when you will inform people at the beginning allows you to establish responsibility before the work moves between procurement, technical teams, communications and compliance. It also gives you time to distinguish between different forms of AI involvement. Using AI to correct grammar is not the same as using it to recommend an outcome, generate public information or interact directly with someone accessing a service. Your disclosure should help people understand the role AI played, not merely announce its presence.
There is now a regulatory reason to examine this carefully. The European Commission has published guidance on the transparency obligations under Article 50 of the EU AI Act, which apply from 2 August 2026. The guidance covers areas including direct interaction with AI systems, machine-readable marking of synthetic content, and the labelling of deepfakes and certain AI-generated public-interest texts.
Even where the legislation does not apply directly to your organisation, the guidance offers a useful place to begin. It encourages you to treat transparency as part of system design and organisational responsibility, rather than a disclaimer added immediately before publication.