BlogAI voice agents

Should an AI voice agent say it is an AI?

Should an AI voice agent say it is an AI? Where disclosure is required as of October 2026 (US, California, Utah, EU), why to do it anyway, and wording to use.

By the DialSavvy team6 min read

Fig. 1 · An agent hands a caller to a rep · Concept · fictional data
On this page

Yes. An AI voice agent should say it is an AI in its first sentence, on every call, along with the name of the business it calls for. As of October 2026, no US federal rule requires the words “I am an AI”, but some states and the EU do in some cases, and US rules already require every AI-voice call to identify the business at the start.

What the rules require, as of October 2026

AI voice agent disclosure rules come from 2 directions: rules for artificial-voice calls, and newer rules for AI systems that talk with people. Since the FCC’s declaratory ruling FCC 24-17 (8 Feb 2024), an AI-generated voice counts as an “artificial or prerecorded voice” under the US Telephone Consumer Protection Act, so the artificial-voice rules apply to AI agents.

Where Rule What it requires Status
US, federal 47 CFR 64.1200(b) At the start, the business’s registered name; a phone number during or after the message; for sales calls to mobile and home numbers, an automated opt-out within 2 seconds of the identification In force
US, federal FCC 24-84, proposed rule Would define AI-generated calls and require callers to disclose AI use at the beginning of each call and when they ask for consent Proposed 10 Sep 2024; no final rule
California Public Utilities Code 2874 Calls from an automatic dialing-announcing device must open with a live, natural-voice announcement that names the business, asks if the person consents to hear the message, and says if the message uses an artificial voice In force since 1 Jan 2025
California Business and Professions Code 17941 No using a bot “online” to mislead people about its artificial identity to drive a sale; disclosing that it is a bot is a defense In force since 1 Jul 2019
Utah SB 226 (2025), as summarized by the Future of Privacy Forum A supplier using generative AI with consumers must disclose it when the person clearly asks; licensed professions must prominently disclose it in high-risk interactions In force since 7 May 2025
European Union AI Act Article 50(1) AI systems that talk directly with people must be designed so the people are told they are interacting with an AI, unless it is obvious, at the latest at the first interaction Applies from 2 Aug 2026

A few notes on reading the table:

  • The FCC proposal is stalled, not closed. The FCC’s regulatory agenda published on 14 Aug 2026 lists the proceeding on AI-generated calls under long-term actions, with the next action “undetermined”.
  • California’s 2 laws have limits. Section 2874 covers automatic dialing-announcing devices, which section 2871 defines as equipment that dials stored or generated numbers and plays a prerecorded message, and section 17940 defines “online” as websites and apps. Whether either reaches a conversational AI agent on a phone call is a question for your counsel.
  • The EU rule falls on the provider that builds the AI system. A business that builds its own agent may be the provider.

Why say it on every call, even where no rule requires it

  • The rules differ by place and are still moving. One opening line for every call is simpler than tracking each state and country, and it matches what the FCC proposed in 2024.
  • Your consent wording already says it. In the US, prior express written consent for sales calls must authorize calls made with an autodialer or an artificial or prerecorded voice. Saying so on the call matches what the person agreed to.
  • People will ask. Utah requires a truthful answer when a person clearly asks, and California’s bot law targets bots that mislead about being artificial. An agent that dodges the question creates risk on a recorded call.
  • The hand-off to a rep is cleaner. A person who knows they spoke with an AI is not surprised when a rep joins with notes from the conversation.

Sample wording you can adapt

Replace the brackets with your details, and have your counsel review the final script for the places you call.

Outbound callback, consent on file: “Hi, this is an AI assistant calling for [registered business name]. To stop these calls, say ‘stop calling’ or press 9. You asked for a call about [request] on our website [when]. You can ask for a person at any time, and you can reach us at [phone number].” On a sales call, the opt-out comes within 2 seconds of the identification.

Inbound call: “Thanks for calling [business name]. I’m an AI assistant, and this call is recorded. I can answer questions, book a time, or put you through to a person.”

When someone asks if it is a person: “No, I’m an AI assistant for [business name]. Would you like me to connect you to a person?”

Voicemail on a sales call: “This is an AI assistant calling for [business name] about your request for [request]. Please call us back at [phone number]. To stop these calls, call [toll-free opt-out number].” In the US, an artificial-voice sales message left on voicemail must give a toll-free number that reaches the automated opt-out.

A checklist for your agent’s opening

  1. Name the business in the first sentence, using its registered name.
  2. Say it is an AI in the same sentence. If the agent has a first name, keep “AI assistant” next to it.
  3. Say the call is recorded, if it is. Recording consent rules vary by state and country.
  4. Offer a person. Tell the caller they can ask for one at any time.
  5. Offer the opt-out on sales calls: an automated voice or keypress option within 2 seconds of the identification.
  6. Give a phone number the person can call back.
  7. Answer “Are you a real person?” truthfully, every time it is asked.
  8. Keep the record: the recording or transcript shows what the agent said.

Disclosure is half of the rules. The other half is consent before the call: reps make cold calls; AI agents call people who agreed to it. See AI voice agent consent rules, and for callbacks to new leads, speed to lead.

How DialSavvy handles this

DialSavvy Agents are in early access: we build your first agent with you. Each agent will say it is an AI before anything else and tell the caller the call is recorded. A person will always be one step away: the agent will put the caller through to a free rep or take a message. Checks for AI-voice consent before an outbound agent dials are coming. Agents will not clone your reps’ voices. See DialSavvy Agents and how compliance checks work for reps today.

Write your agent’s first 2 sentences this week, check them against the list above, and ask your counsel to review them for each state and country you call.

Sources

  1. 47 CFR 64.1200, Delivery restrictions, eCFR (ecfr.gov)
  2. Implications of Artificial Intelligence Technologies on Protecting Consumers from Unwanted Robocalls and Robotexts, Declaratory Ruling (FCC 24-17), Federal Communications Commission (docs.fcc.gov)
  3. Implications of Artificial Intelligence Technologies on Protecting Consumers From Unwanted Robocalls and Robotexts, proposed rule (FCC 24-84), 89 FR 73321, Federal Register (govinfo.gov)
  4. FCC Semiannual Regulatory Agenda, Unified Agenda of Federal Regulatory and Deregulatory Actions 2026 (14 Aug 2026), Federal Register (federalregister.gov)
  5. California Public Utilities Code section 2874 (as amended by AB 2905, effective 1 Jan 2025), California Legislative Information (leginfo.legislature.ca.gov)
  6. California Public Utilities Code section 2871 (automatic dialing-announcing device), California Legislative Information (leginfo.legislature.ca.gov)
  7. California Business and Professions Code sections 17940-17943 (Bots), California Legislative Information (leginfo.legislature.ca.gov)
  8. Overview of Utah's 2025 Enacted AI Legislation (April 2025), Future of Privacy Forum (fpf.org)
  9. AI Act, Article 50: Transparency obligations for providers and deployers of certain AI systems, European Commission, AI Act Service Desk (ai-act-service-desk.ec.europa.eu)
  10. AI Act, Article 113: Entry into force and application, European Commission, AI Act Service Desk (ai-act-service-desk.ec.europa.eu)

Checked against these sources by the DialSavvy team. How we write. Spotted an error? Email hello@dialsavvy.com.

Questions

Anything else, ask in your demo.

Book a demo

Is it illegal for an AI to call without saying it is an AI?

As of October 2026, no US federal rule requires the words "I am an AI", but every AI-voice call must name the business at the start. Some states have their own rules: California requires calls from automatic dialing-announcing devices to say when a message uses an artificial voice. Check the states you call with your counsel.

Does the EU AI Act require AI phone agents to disclose?

Article 50(1) requires providers to design AI systems that talk directly with people so that those people are told they are interacting with an AI, unless it is obvious. It applies from 2 Aug 2026, and the information must be given at the latest at the first interaction.

What should an AI agent say when someone asks if it is a person?

The truth, in one sentence: "No, I'm an AI assistant for [business]. Would you like me to connect you to a person?" Utah requires a supplier to disclose generative AI when a person clearly asks, and California's bot law targets bots that mislead people about being artificial.

Does saying it is an AI replace consent?

No. Disclosure happens on the call; consent comes before it. In the US, an AI-voice sales call to a mobile or home number needs the person's prior express written consent, whatever the agent says once the call connects.

Keep reading

Put an AI agent on your phones.

Early access is by invitation. We set up your workspace, numbers and caller ID, then send your invite.