September 24, 20269 min read

AI Agents That Make Phone Calls for You: How Instinct and Meta Muse Calling Work, and Is It Legal?

Instinct and Meta Muse can now phone businesses for you, and Meta had humans make some calls. How AI calling works, the disclosure rules, and how to tell.

ByLokesh Kumar· Founder & Builder, Vyra
Share on XShare on LinkedIn

The short answer

Yes, AI agents can now phone businesses for you. Instinct and Meta's Muse both added calling in September 2026, joining Google's AI calling in Search. No rule we reviewed clearly bans using one to call a business, but AI voices fall under US robocall law, the EU now requires disclosure that you're talking to an AI, and recording rules vary by state. Meta also briefly routed some Muse calls to human contractors.

This isn't legal advice. It's a plain-English guide to what these agents do, the rules that apply, and what to watch for on either end of the call.

Which AI agents can make phone calls?

Three products can now place calls for you, and one of them has been doing it for years. A fourth, will.i.am's FYI.AI, gives its agents their own phone numbers, though we couldn't verify its availability (Axios).

AgentWho it callsAvailabilitySays it's AI?Humans involved?
Google AI calling in SearchLocal businessesAll US Search users since Jul 2025; expanding in 2026Google committed to built-in disclosure in 2018Duplex could hand hard cases to human operators
Meta MuseVerified US businessesBeta, expanded 16–17 Sep 2026Not detailed in Meta's materialsSome calls placed by human contractors in a test, reportedly rolled back
Instinct ("Instinct Concierge")Businesses (restaurants, dentists, cable companies)Early access for select usersNot stated in any source we foundNot stated

Sources: Google I/O 2026; Google Keyword blog; Google Research; TechCrunch; Axios.

Instinct

Instinct connects to a user's email, messages, screen, audio and location, and can proactively text or call the user (Axios). On 17 September its founder Noah Shinn announced "Your Instinct can now handle phone calls." The examples were practical: booking a restaurant that doesn't take online reservations, getting on a dentist's cancellation list, fixing a cable bill (TechCrunch).

Meta Muse

The same week, Meta said it had "expanded the Muse beta for outbound calls to U.S. businesses," starting with users who asked for it (quoted by TechCrunch). Muse itself launched on 8 September; its launch post doesn't mention calling (Meta Newsroom).

Google, the one that started it

Google demonstrated Duplex in May 2018, an AI that could book reservations by phone (Google Research). The demo voice sounded human and never said it was AI, and after the backlash Google said it was "designing this feature with disclosure built-in" (CBS News). Today, Google Search can call local businesses for you across the US, and businesses can opt out (Google Keyword blog).

Was it really an AI making the call?

Not always, it turned out. 404 Media reported that Meta added a "human agent layer" in which some Muse calls were handed to trained people in call centers, and some users only learned a human had made their call afterwards.

Reuters reported further details, repeated by several outlets because the original is paywalled. The human option was switched on for about half of Meta's employees in the test. A Meta vice president said launching it without proper disclosure "was a miss" and that Meta "rolled the feature back for now." A Meta spokesperson said it would only launch publicly "with appropriate disclosures and safeguards" (TechSpot; The Next Web).

The same reporting said one contractor made a racist remark on a call negotiating an employee's internet bill, and that employees had warned sensitive data could reach contractors. Meta said that contractor would no longer work on its projects (TechSpot).

The lesson isn't that human help is wrong. Google designed Duplex to hand hard cases to human operators too (Google Research). It's that when you hand an agent a task involving your bills or health, you should know who actually handles your details.

We found no rule that clearly bars using an AI agent to call a business on your behalf, but several rules shape how it has to be done. These are the main ones.

US: AI voices count as "artificial" under robocall law

In February 2024, the FCC ruled unanimously that calls using AI-generated voices are "artificial" under the Telephone Consumer Protection Act (TCPA), effective immediately. That means robocall rules apply, including needing "prior express written consent before robocalling consumers" (FCC).

The TCPA restricts artificial-voice calls to any cell phone number, including a business's mobile, and to residential lines, unless the called party has consented (47 U.S.C. § 227). Calls to a business landline are treated differently. Since many small businesses answer on mobiles, whether a given AI call is covered depends on facts a lawyer would need to assess.

US states: disclosure laws

EU: you must be told you're talking to an AI

Under Article 50 of the EU AI Act, people must be informed that they're interacting with an AI system unless that's obvious, at the start of the first interaction. This obligation has applied since 2 August 2026, and fines can reach €15 million or 3% of worldwide turnover (European Commission). Google describes its AI calling as US-only (Google Business Profile Help), and Meta's beta covers US businesses (TechCrunch).

Recording the call

US federal law allows recording when one party consents (18 U.S.C. § 2511). Some states, including California, require every party to consent before a confidential call is recorded (California Penal Code § 632). That matters because Google says its automated business calls "are monitored and recorded for quality assurance" (Google Business Profile Help). Muse and Instinct haven't publicly said whether they record or transcribe calls.

How do you know if you're talking to an AI on the phone?

You can't always tell from the voice anymore, so rely on behavior and ask directly:

  1. Ask: "Am I speaking with an AI?" A well-designed agent should answer honestly, and in the EU an AI system has to disclose itself unless that's obvious.
  2. Listen for the opener. Google committed in 2018 to having its calling AI identify itself. A caller that won't say who or what it is deserves caution.
  3. Watch for narrow scope. Calling AIs are built for specific tasks such as bookings, prices and business hours (Google Research), so a caller that sticks rigidly to one request may be automated.
  4. Don't give payment or ID details to an unexpected caller, human or AI. Call the business back on a number you trust.

If you run a business, check whether you can opt out. Google lets businesses decline its AI calls during a call or in Business Profile settings (Google Business Profile Help).

Should you let an AI agent make calls for you?

For simple, low-stakes tasks, such as checking a price, booking a table or joining a waitlist, it's a genuine time-saver. For anything involving money, health or your identity, apply the same caution you'd use for giving an AI agent access to your email: know exactly what it can say and agree to on your behalf, and make it check with you before committing.

A call is an irreversible action, which puts it squarely in the blast radius of whatever access you've given the agent. Once the agent has agreed to a price or shared your details, you can't un-say it. That's why the most important feature to look for is an approval step before the agent commits to anything, the pattern we describe in human-in-the-loop AI agents. For a wider view of how these assistants compare, see the personal AI agent race of 2026.

Common questions about AI agents making phone calls

We found no rule that clearly bars it, but rules apply. In the US, AI voices count as "artificial" under the TCPA, which restricts such calls to cell phones, including business mobiles, and residential lines without consent. Some states require disclosure when asked, and the EU requires telling people they're talking to an AI. This isn't legal advice.

Does an AI calling agent have to say it's an AI?

In the EU, yes, since 2 August 2026, unless it's obvious. In the US it depends on the state: Utah requires businesses using AI to disclose it if a consumer asks, and California's bot law bans deceptive bots in some contexts. Google committed in 2018 to building disclosure into its Duplex calling AI.

Can an AI agent call to book a reservation or appointment for me?

Yes. Google Search can call local businesses for you across the US, Meta Muse is testing calls to verified US businesses, and Instinct offers calling in early access. Tasks like reservations, pricing and cancellation lists are the typical use.

Can a business block AI calls?

Businesses can opt out of Google's AI calls during a call or through Google Business Profile settings. We found no published opt-out for Muse or Instinct calls.

Sources

The more an agent can do in the real world, the more it matters that it asks first. Vyra by Vyraagi is being built with a check that flags irreversible actions before they run. If that's what you want in an agent, join the waitlist.

Vyra is in closed alpha now, with a Founders Beta ahead of public launch.

Join the WaitlistSee Plans

Related reading

Autonomous Goal Tracking: How OKR-Driven AI Agents Work in the Background

How an AI goal engine turns a stated objective into key results and tasks, then advances them on its own — the architecture behind autonomous goal tracking.

Is Meta Muse Safe on Your Mac? What It Can Access, the Zero-Day, and How to Lock It Down

Meta's Muse agent can act across files, Messages, Mail and Calendar on a Mac. What it can reach, the dictation zero-day Meta says it fixed, and how to limit it.

What actually happened in September 2026's rogue AI agent incidents: a sourced timeline

A sourced timeline of the OpenAI agent wiki board, RubyGems packages, urlquery.net scans and the Medicare portal breach: confirmed, claimed, unknown.