Does Article 50 apply to copilot-embedded products?

"Copilot" covers two situations that the EU AI Act treats very differently, so start by asking which one you're in.

You ship a product with an AI assistant in it. If your SaaS or app exposes a chat-style AI helper to your users under your product's brand (even if it's Azure OpenAI, Microsoft Copilot Studio, or another vendor's stack underneath), you are generally the provider of that AI system under the Act. The Article 50(1) duty to tell users they're interacting with AI is yours. That holds for B2B products too: the people typing into your copilot are natural persons, and the Act's duty runs to them, not to the corporate entity that signed your contract. (A narrow exemption exists for AI that's obvious from context; a code assistant used only by developers is the draft guidelines' example. A general-purpose in-app assistant shouldn't assume it qualifies.)

You merely use Microsoft 365 Copilot (or similar) inside your company. Microsoft-branded Copilot used by your own staff, unmodified, leaves you a deployer of Microsoft's system, and the provider-side disclosure duty for it stays with Microsoft. Whether Article 50(1) reaches purely internal audiences at all is a genuinely unsettled question (see the internal-chatbots page), but it isn't the configuration regulators are expected to check first.

For the product case, compliance is cheap and concrete: an AI notice at or before first interaction with the assistant, plus a persistent indicator (a visible "AI" label on the copilot panel does this job well), accessible to screen readers. A mention in your documentation or terms alone is expressly insufficient under the draft guidelines.

Also worth documenting: whether your copilot merely answers, or also generates content your users then publish. The second raises separate marking questions under Article 50(2), a provider-side duty with its own page.