Updated 2026-07-29 · 4 min read
First, who this binds. What follows is European Union law: it applies to practices in the EU, and to anyone placing an AI assistant on the EU market. If your practice is in the United Kingdom or the United States, no equivalent obligation is in force for you today. It's still worth reading, because in a healthcare setting the standard below is one you'd want to meet whether or not a regulator is asking.
There's a date marked for any European business using an artificial intelligence assistant: 2 August 2026. From that day article 50 of Regulation (EU) 2024/1689 — the «EU AI Act» — becomes enforceable, and its central idea fits in one sentence: nobody should be talking to a machine believing they're talking to a person.
If an assistant answers your practice's WhatsApp or phone, this concerns you directly. And in a healthcare setting, where trust is everything, meeting it properly isn't just an obligation: it's a calling card.
What the rule says, without the jargon
Article 50(1) requires that anyone interacting with an AI system knows it from the very first moment, clearly and perceptibly. Hiding it in the privacy policy doesn't count: it has to be in the conversation itself. In practice: the first message of a new WhatsApp conversation has to say that an artificial intelligence assistant is answering, and the phone greeting has to say it too, before anything else.
An important nuance: the obligation falls on the provider of the system, not on the practice. A serious assistant puts the notice up itself in every new conversation and keeps a record of when it sent it — that traceability is the difference between «we think we told them» and being able to prove it to whoever asks.
Free tone, obligatory content
The rule doesn't dictate wording: «You're speaking to Sonia, the artificial intelligence assistant at Sonrisa Dental Practice» complies perfectly and sounds like your practice. What you can't do is drop the essential part. The right balance is freedom of tone with obligation of content: you choose your assistant's name and style, and the system checks that the text still says what the law requires.
Two details give away a well-built system: on the phone the long form is used — «artificial intelligence», because initials read out by a voice engine sound like anything — and the assistant can't share a name with somebody on the team. If your hygienist is called Lucy and your assistant is too, your patients will never know who they spoke to.
In a practice, transparency is an advantage
The usual objection is commercial: «if I say it's an AI, they'll hang up». Experience says the opposite: what burns trust isn't the notice, it's discovering the trick halfway through the conversation. In a healthcare setting this weighs double — a patient who feels deceived at reception will doubt the rest of the practice too.
The notice, well written, frames the conversation: somebody who knows they're talking to an assistant asks more directly, understands the occasional limitation, and values the «let me pass you to the team» when it's genuinely needed. The assistant sorts and books; it never assesses symptoms and never diagnoses — and saying so from the first second is part of doing it properly.
The short list to be in order
First: the notice in the first message of every new chat conversation, placed by the system and not by anybody's memory. Second: the phone greeting, with «artificial intelligence», the assistant's name and the practice's. Third: a record of every notice sent. Fourth: validation of the text if you personalise it — free tone, untouchable content. And fifth: a clear identity, never passing as human and never sharing a name with the team.
If your current system can't show you those five points working, 2 August 2026 is a good reason to look at it again — and outside the EU, four of the five are simply what an honest assistant does. With ClaudIA it's already done: the notice goes out on its own, it's recorded, and all you choose is how it sounds.