Does Your AI Receptionist Have to Say It's an AI? UK, EU and US Rules for Small Businesses

A caller rings a dental practice at quarter to eight in the evening. A warm voice answers, offers Thursday at 10, takes their name and date of birth, and books the appointment. The caller hangs up thinking they spoke to the receptionist. Should the voice have said it was an AI?

It depends on where your callers are. In the EU, the AI Act has required since 2 August 2026 that AI systems which talk directly to people are designed to tell them they are dealing with AI, unless that is obvious. The UK has no AI-specific disclosure law, but the competition regulator expects you to tell callers whenever it might affect their decisions, and data protection law already asks you to explain yourself when you collect their details. In the US, the answer turns on the state and on whether you are answering calls or making them. In all three places, a clear sentence at the start of the call is the safe default.

This is general information, not legal advice. If you work in a regulated sector such as healthcare, financial services or legal services, check the rules for your sector with a qualified adviser.

EU: Article 50 of the AI Act

Article 50(1) of the EU AI Act says providers must ensure that AI systems intended to interact directly with people are designed so that "the natural persons concerned are informed that they are interacting with an AI system". The exception is where this is obvious to a person who is "reasonably well-informed, observant and circumspect", plus a narrow one for systems authorised by law to fight crime. Article 50(5) says the information must be given "in a clear and distinguishable manner at the latest at the time of the first interaction", and must meet accessibility requirements.

The rule has applied since 2 August 2026. The Digital Omnibus on AI, published in the Official Journal on 24 July 2026, did not change it. The omnibus gave a grace period to 2 December 2026 only for the separate duty to mark AI-generated content in a machine-readable way, and only for systems already on the market before 2 August 2026. The Commission's Article 50 FAQ confirms that this is the only grace period.

The duty sits with the provider: the company that develops the AI system, or has it developed, and puts it on the market or into service under its own name or trade mark. If you buy an AI receptionist from a vendor, the vendor is normally the provider and you are the deployer. A business that builds a system in-house, or commissions one, and uses it under its own name is a provider too. In our reading, the same may apply if you assemble your own voice agent on a platform and run it under your business's name. Check this with your vendor and adviser.

The Commission's guidelines on Article 50, dated 20 July 2026, are specific about phones. They are not legally binding, but they show how the Commission reads Article 50. For voice and telephony they point to "explicit spoken statements at the beginning of the interaction", with reminders in longer calls, and say tones and other audio cues "are not considered sufficient by themselves". They read the "obvious" exception narrowly, limiting it to cases with "almost no doubt left". A human-sounding voice, or a public audience that includes elderly people or people with disabilities, makes the exception harder to rely on. AI agents that make bookings for someone should disclose both that they are AI and on whose behalf they act.

The Act also reaches providers and deployers outside the EU "where the output produced by the AI system is used in the Union". If you serve EU customers, ask your vendor how its system meets Article 50. Occasional EU callers are a greyer area: the guidelines say incidental or unforeseeable use alone should not trigger the obligations for a non-EU provider whose system is not on the EU market. Breaches of Article 50 can be fined up to €15 million or 3% of worldwide annual turnover, whichever is higher. For SMEs and, since the omnibus, small mid-caps, the lower of the two applies. These are maximums, and Member States set the penalty rules.

UK: no AI disclosure law, but consumer and data protection law still apply

The UK has no AI-specific statute. A House of Commons Library briefing published on 14 October 2026 puts it plainly: "The UK does not have any AI-specific regulation or legislation covering AI as a technology." AI is regulated through the existing law that applies where it is used.

For a receptionist, consumer law matters most. In March 2026 the Competition and Markets Authority published guidance on using AI agents. It says you are responsible for what an AI agent does "in the same way you are responsible for what an employee does", even when someone else designed or provides it. On disclosure it says: "if the fact they are dealing with AI rather than a person might affect people’s decisions then you should tell them." It warns against overstating what the AI can do and notes fines of up to 10% of worldwide turnover for breaking consumer law.

The legal test is in section 225 of the Digital Markets, Competition and Consumers Act 2024. A commercial practice is unfair if it is likely to cause the average consumer to take a transactional decision they would not otherwise have taken, for example through a misleading action or a misleading omission. In our reading, the clearest risk is a bot that says "yes" when a caller asks whether they are speaking to a person. Staying silent is less clear-cut, which is why the CMA frames disclosure around whether it might affect decisions. These rules protect consumers, so a supplier calling your trade line is outside them.

Data protection gives a second reason to speak up early. The ICO's guidance on the right to be informed says you "must provide privacy information to individuals at the time you collect their personal data from them", and suggests layering: a short notice now and fuller details elsewhere. An AI receptionist that takes names, numbers and appointment details collects personal data from the first minute, so the opening of the call is where the short notice belongs. The ICO marks this guidance as under review following the Data (Use and Access) Act.

If you record calls, say so in the same opening. Our guide to AI call answering for tradespeople covers the ICO's position on recording and what to ask a provider about storage.

Calls you make are stricter. The ICO's telephone marketing guidance says you must not make an automated marketing call, "a call made by an automated dialling system that plays a recorded message", without specific consent to that type of call. That guidance does not say whether a conversational AI caller counts, so our view is to treat AI marketing calls as needing consent. Separately, Ofcom's persistent misuse policy treats silent and abandoned outbound calls as misuse. When it assesses abandoned calls, it considers whether an information message identifying the caller was played, and whether it started within two seconds.

US: federal rules cover calls you make, states fill the gaps

The federal rules we found apply to calls you make, not calls you answer. In a February 2024 declaratory ruling, the Federal Communications Commission confirmed that the Telephone Consumer Protection Act's restrictions on "artificial or prerecorded voice" calls cover "current AI technologies that generate human voices". Unless an exemption applies, such calls need the called party's prior express consent, or prior express written consent for telemarketing. Under 47 CFR 64.1200(b), every artificial or prerecorded voice message must, at the beginning, "state clearly the identity of the business" responsible and give a phone number. That is a duty to say who is calling, not to say the voice is AI.

California's bot disclosure law (SB 1001, operative since July 2019) makes it unlawful to use a bot to talk to someone in California online with intent to mislead them about its artificial identity and knowingly deceive them about the content of the communication in order to drive a sale, unless you disclose that it is a bot. The law defines "online" as public-facing websites and apps, so it may not reach a phone call. A separate rule, Public Utilities Code section 2874, amended with effect from 1 January 2025, requires a caller using an automatic dialing-announcing device to make a live, unrecorded announcement first, which must tell the person called if the prerecorded message uses an artificial voice. It applies to calls you make.

Utah is more direct. Utah Code 13-77-103 requires a business using generative AI in a consumer transaction to disclose that the person is dealing with AI "and not a human" when the person clearly asks. People in licensed or state-certified occupations regulated by the state's Department of Commerce must disclose up front, "verbally at the start of a verbal interaction", when the interaction is high-risk, such as collecting health or financial data or giving medical, legal or financial advice. Maine's Public Law 2025, chapter 294 bars using an AI chatbot, defined to cover "textual or aural" conversation, in a way that may mislead a reasonable consumer into believing they are talking to a human, unless the consumer is clearly told.

Recording is a separate question. In California, recording a confidential call without the consent of all parties is an offence under Penal Code section 632; other states differ. State law on AI is changing, so check the current position wherever your callers are.

UK, EU and US compared

What to say at the start of the call

• Required for EU-facing systems: a clear statement that the caller is talking to AI, at the start. Your vendor has to build it in. Your part is to keep it in and not bury it.
• Required in the EU and some US states, and the clearest UK consumer-law risk: never let the AI claim to be a person. In the EU it would defeat Article 50(1). Utah requires a truthful answer when a caller asks. Maine bars misleading consumers into thinking they are talking to a human. In the UK a false "yes" is the clearest consumer-law risk.
• Required in the UK and EU when you collect personal data: say who you are, why you are collecting details and where the full privacy notice is. Say if you record.
• Good practice: name the business the AI works for, offer a person, and say when a person takes over.

An opening that covers all of this can be short. For example: "Thanks for calling Northside Dental. I'm the practice's AI assistant. Calls are recorded to manage bookings, and our privacy notice is on our website. I can book or change appointments, or put you through to the team." When Evoogen sets up call answering for a client, this opening is one of the first things we agree, because it depends on your sector, your recording settings and where your callers are.

Answer the question "Am I talking to a robot?" truthfully every time, wherever the caller is. If a person joins the call, the caller should hear that the AI has handed over. The Commission's guidelines suggest repeating the AI notice when the system's role changes during a call.

A worked example (hypothetical)

Assumptions. A physiotherapy clinic has two sites: one in Liverpool (England, UK) and one in Dublin (Ireland, EU). It uses a voice AI receptionist from a vendor that sells the product under its own brand. The AI answers inbound calls only, books appointments and records calls for quality checks. Some callers are elderly. The clinic makes no outbound marketing calls. These details are invented for illustration.

• Dublin line. Article 50 applies. The vendor is the provider and must build in the disclosure. The voice sounds human and some callers are elderly, so the clinic cannot treat the AI as obvious. It asks the vendor to confirm in writing how the system meets Article 50 and checks that the greeting has not been changed to hide it.
• Liverpool line. No AI-specific statute applies. Under the CMA guidance the clinic is responsible for the AI as if it were an employee. Callers may describe symptoms or ask about prices, so knowing they are talking to AI could affect their decisions. The clinic tells them.
• Both lines. Callers may describe symptoms, and health data gets extra protection in data protection law (in the EU, under Article 9 of the GDPR), so the clinic takes sector-specific data protection advice. The same opening runs on both lines: it names the clinic, says the assistant is AI, explains recording, points to the privacy notice and offers a person.

The main extra cost is one more sentence per call and a written answer from the vendor.

Common misconceptions

• "If the voice sounds natural, nobody needs to know." The Commission's guidelines treat a human-sounding voice as a reason the AI is less obvious, which makes disclosure more necessary, not less.
• "The vendor handles compliance." In the EU the vendor carries the design duty. In the UK the CMA says you are responsible for what the agent does, even when a third party provides it.
• "The FCC ruling makes AI receptionists illegal in the US." The ruling says calls made with an AI-generated voice need the called party's consent. It does not address calls your customers make to you.
• "A beep is enough." The guidelines say audio cues alone are not sufficient for telephony.
• "The EU delayed the AI Act." The omnibus delayed the high-risk rules and gave a short grace period for content marking. Article 50(1) applied on 2 August 2026 as planned.

What good looks like

• The first sentence says the caller is talking to an AI assistant, names your business and mentions recording.
• The AI answers "Are you a person?" honestly, and a caller can reach a person easily.
• Your vendor has confirmed in writing how its system meets Article 50 and where call data are stored.
• Someone reviews transcripts and complaints regularly, as the CMA expects for AI agents.

The bottom line

For EU-facing systems, telling callers they are talking to AI is now a legal requirement. In the UK it is what the competition regulator expects whenever it could affect a caller's decision, and data protection law already asks you to explain yourself at the start. In the US, some states require it in some situations, and it is a safe default in the rest. Making it the first sentence on every line deals with the disclosure question in all three.

If you want an AI receptionist set up with the right opening, recording notice and hand-off for your callers, Evoogen can map your call flow and implement it. See how Evoogen can help with calls and bookings.