A checklist for the EU AI Act if you built something with GLACIER

Where things stand: the EU AI Act's transparency rules are already in force, and they apply whether your AI system is high-risk or not. Crucially, these duties apply to anyone using an AI system in something they built, not only to the company that built the underlying AI model. If you built a chatbot, an AI companion, or an automated content tool with GLACIER, you are very likely covered by these rules today, not at some point in the future.

The four duties that apply, and to whom:

You must tell users they're interacting with AI unless it's already obvious from context. AI-generated audio, image, video and text needs machine-readable marking — tools already on the market before these rules took effect have a short grace period to add this if it isn't there yet. If your product uses emotion recognition or biometric categorisation on users, you must disclose that to the people it's applied to — this responsibility sits with you as the one using the feature, not with the tool that built it for you. And if you publish AI-generated content on a matter of public interest without a human editor reviewing it first, you need to disclose the AI origin.

The penalty for getting this wrong is substantial — among the highest in the regulation, calculated as a percentage of global turnover, not a fixed fine.

A five-minute checklist:

If you're unsure whether a specific feature you built with GLACIER is covered, write to [email protected] and we'll help you work through it.