Do I Have to Tell Customers They're Talking to an AI? Disclosure Rules for Business Phone Agents in 2026
Every legal claim below is attributed to a published source read on 2026-10-10. This is a business guide, not legal advice — check with a lawyer for your jurisdiction. No invented statistics, no client claims.
Short answer: Yes, in most cases. Since August 2, 2026, the EU's AI Act requires AI systems that interact with people to disclose they're AI, and US law treats AI-voice calls as regulated calls requiring caller identification and (for outbound calls) prior consent. The safe move everywhere: have your AI agent introduce itself as AI in its first sentence. It's a three-second fix.
The 2026 rules that actually matter
Four rules cover nearly every business deploying an AI receptionist or voice agent. None of them ban the technology — they all converge on one requirement: don't let a customer believe they're talking to a human when they aren't.
1. EU: AI Act Article 50 — in force since August 2, 2026
Article 50 of the EU AI Act (Regulation (EU) 2024/1689) applies to any AI system that interacts directly with a natural person — not just high-risk systems. The rule: the system must be designed so the person is informed they are interacting with AI, unless that is obvious from the context. AI-generated audio must also be marked in a machine-readable way; systems already on the market before August 2, 2026 got until December 2, 2026 to meet the marking sub-obligation. Fines run up to €15 million or 3% of worldwide annual turnover (reported by Cloud Security Alliance and technology.org, 2026). In practice, your AI receptionist's first sentence must name itself as AI for EU callers.
2. California: AI Transparency Act — operative August 2, 2026
California's AI Transparency Act (SB 942, signed September 2024, amended by AB 853 in October 2025) applies to "covered providers" — publicly accessible generative AI systems with more than 1 million monthly users in California. They must offer a free AI-detection tool, give users a visible "manifest" disclosure option, and embed "latent" (machine-readable) disclosures in AI-generated image, video, and audio. Civil penalties: $5,000 per violation, with each day counted separately (reported by Jenner & Block; Enterprise DNA, 2026).
The honest nuance: this law binds the big AI providers, not your dental clinic or real estate agency directly. But it pushes your vendors toward disclosure-by-default — ask your AI receptionist vendor how they comply, and put the answer in writing.
3. California's bot-disclosure law — on the books since 2019
The B.O.T. Act (SB 1001, operative July 1, 2019) makes it unlawful to use a bot to interact with a Californian with the intent to mislead about its artificial identity in order to incentivize a commercial transaction — unless the disclosure is "clear, conspicuous, and reasonably designed to inform" the person it's a bot. Violations can draw up to $2,500 each under the state's Unfair Competition Act (reported by Epiq Global; Financier Worldwide).
It was written for online bots, not phone agents — but it's the closest US state law on AI-identity deception, and it shows where American regulators have been heading for seven years: disclosure up front.
4. US federal: the FCC's 2024 AI-robocall ruling
On February 8, 2024, the FCC unanimously ruled that AI-generated voices in robocalls count as "artificial or prerecorded voice" under the Telephone Consumer Protection Act (TCPA). That means outbound AI calls require the called party's prior express consent, and AI callers must provide identification and disclosure information plus opt-out methods (reported by the Harvard Journal of Law & Technology; Orrick InfoBytes).
Critical distinction: this governs outbound calls — AI appointment reminders, follow-ups, reactivation campaigns. Your inbound AI receptionist answering your own business line is a different scenario. But the moment your agent dials out, TCPA applies.
Google set the cultural precedent back in 2018: Duplex calls disclosed they came from "an automated system" upfront (reported by Search Engine Land). The 2026 rules just made that instinct into law.
Inbound vs. outbound: what you must actually do
| Inbound AI receptionist (answers your line) | Outbound AI calls (reminders, follow-ups) | |
|---|---|---|
| Disclosure | Introduce as AI in the first sentence — required in the EU, expected everywhere else | Identify the business and that it's AI; US caller-ID rules apply |
| Consent | The caller chose to call you; no extra consent step | Prior express consent required in the US (FCC/TCPA, 2024) |
| Biggest risk | Customer feels deceived → complaint, bad review, regulator attention | TCPA lawsuits — reported statutory damages of $500–$1,500 per call |
The five-line compliance script
- Open with the name, not a dodge. "Hi, I'm Ava, the AI assistant at [Business]. How can I help?" Three seconds, done — and it satisfies the EU's first-interaction requirement.
- Re-disclose on transfer or escalation. If the call moves to a sensitive topic or a human takes over, say so.
- Always offer the human path. "Want a person instead? I'll connect you now." Ofcom's September 2026 assessment of AI in telecoms stressed exactly this: customers must be told clearly when they're dealing with AI, and AI must not become a barrier to reaching a person (reported by BroadbandSwitch, Sept 25, 2026).
- Log consent for anything outbound. Keep consent records and call logs — TCPA disputes are won on documentation.
- Don't confuse AI rules with recording rules. Call-recording consent (one-party vs. two-party states) is a separate legal axis. Clear both.
FAQs
Does a small business really fall under the EU AI Act? Article 50 applies to AI systems on the EU market regardless of company size — and as the deployer facing the caller, the disclosure duty lands on you in practice. If you take calls from EU customers, disclose.
What if the AI only answers after hours? Still an interaction. The disclosure goes in the greeting whatever the hour — that's also when miss rates (and caller frustration) are highest.
Do website chatbots need the same disclosure? Yes. Article 50 covers text chat too, and California's bot law was written for online interactions. Same fix: a visible "AI assistant" label or an opening line that says it.
Won't saying "I'm an AI" hurt conversions? The honest answer: no published A/B data exists either way. What 2026 practitioner guides consistently report is that hang-ups trace to bad configuration — slow replies, looping scripts, no path to a human — not to disclosure. And a customer who discovers a hidden AI trusts you less than one you told upfront.
The bottom line
Disclosure isn't a compliance project; it's one sentence in the greeting. The August 2026 rules made it a legal duty in the EU and the industry standard everywhere else. Businesses that disclose get the efficiency of AI answering with none of the trust risk of a customer feeling tricked. Write the sentence once, and the phone keeps working for you.
VisionQuantech builds AI receptionists with disclosure baked into the greeting — first-sentence AI identification, human-handoff paths, and consent logging for outbound calls. Email [email protected] to talk about a pilot.