The EU AI Act is a sprawling piece of legislation, but for most AI companion operators one set of obligations matters most: transparency.
Under the Act's transparency rules, providers of AI systems that interact directly with people must ensure users are informed they are dealing with an AI — unless that is obvious from the context. Rules on marking AI-generated images, audio and video also apply.
The design tension
For a productivity assistant, disclosure is trivial. For a companion app — a product whose entire value proposition is emotional realism — it is a design problem. How often should a user be reminded? Does a disclosure at sign-up count, or does it need to be persistent? What about voice calls, where there is no on-screen label?
The Act leaves room for interpretation, and guidance will continue to evolve. Sensible operators are not waiting for certainty. Common approaches include:
- Clear disclosure at onboarding, in the app store listing and in the companion's profile.
- A persistent but unobtrusive "AI" indicator in the chat interface.
- Audio cues or spoken disclosure at the start of voice calls.
- Watermarking or metadata on generated images and voice clips.
Beyond the EU
Europe is not alone. Several US states have passed or proposed laws that go further for companion chatbots specifically — including periodic reminders that the user is talking to an AI and protocols for users who express thoughts of self-harm. Operators serving a global audience increasingly design to the strictest common denominator rather than maintaining a patchwork of regional builds.
The bottom line
Transparency does not have to undermine the product. Users of companion apps overwhelmingly know they are talking to an AI; what they value is consistency, warmth and memory. Done well, disclosure becomes part of a trust story that differentiates responsible platforms from the rest.
This article is for information only and is not legal advice. See our regulation tracker for the latest status of rules affecting companion platforms.
