Read this first
This guide is general information, not legal advice. Recording and AI laws vary by state, change often, and depend on the facts of each call. Have counsel review your defaults before you ship them to customers.
The practical takeaway is simple, though. Announcing recording and AI use at the start of every call satisfies the strictest common rules, and it costs a few seconds.
Federal law and one-party consent
The federal Wiretap Act allows a person who is party to a call, or who has the consent of one party, to record it, as long as the purpose isn't criminal or tortious. Most states follow this one-party rule.
For a business recording its own calls, one-party consent means the business itself can consent. That's not enough when the other person is in a state that requires everyone's consent.
All-party consent states
The Reporters Committee for Freedom of the Press lists these states as requiring the consent of every party to record a phone call: California, Connecticut, Delaware, Florida, Illinois, Maryland, Massachusetts, Michigan, Montana, Nevada, New Hampshire, Pennsylvania and Washington.
A few have nuances worth knowing:
- Michigan: the all-party rule applies at least to recordings by someone who isn't part of the conversation. Whether it binds participants is less settled.
- Connecticut and Nevada: all-party for phone calls, one-party for in-person conversations.
- Oregon: the reverse. All-party for in-person conversations, one-party for phone calls.
- Washington: the statute treats consent as obtained when one party announces the recording to everyone in a reasonably effective way, and the announcement itself must be recorded.
- Florida: interception is lawful when all parties have given prior consent.
Interstate calls
A platform's businesses call and get calls from everywhere, so the question is which state's law applies when the parties are in different states. Courts haven't answered it uniformly. Some look at where the recording happens, others at where the recorded person is.
California's Supreme Court, in Kearney v. Salomon Smith Barney (2006), applied California's all-party rule to calls that a Georgia office recorded with California clients, even though Georgia allows one-party recording. The court also said a business that tells all parties at the outset that the call is being recorded would not violate the statute.
The safe default for a platform is to treat every call as if the stricter rule applies: play a recording notice at the start of every recorded call, inbound and outbound, and record the notice along with the call.
AI features count too. An AI receptionist that transcribes or summarizes a call is capturing its contents, so plan on the same notice for calls the AI handles as for calls a person records.
AI-generated voices are "artificial" under the TCPA
In a declaratory ruling adopted February 2, 2024 (FCC 24-17), the FCC confirmed that the Telephone Consumer Protection Act's restrictions on "artificial or prerecorded voice" calls cover current AI technologies that generate human voices, including voice cloning.
That means outbound calls using an AI voice need the called party's prior express consent, absent an emergency or exemption, just like a prerecorded robocall. The FCC also noted that existing rules require artificial or prerecorded voice messages to identify the entity responsible for the call, and that telemarketing messages must offer an opt-out.
The ruling is about calls the AI places. An AI answering a call someone else made is a different situation. In August 2024 the FCC proposed rules (FCC 24-84) that would define AI-generated calls, add disclosure requirements, and expressly exclude technology used to answer inbound calls. We couldn't confirm a final order as of this update, so treat those proposals as pending.
State AI disclosure laws
There is no single federal rule requiring an AI to say it's an AI on every call. Several states have their own rules, and the list is growing. The ones most relevant to business calls as of November 2026:
- Maine (10 M.R.S. section 1500-DD, 2025): a business may not use an AI chatbot, defined to include audio conversation, in a way that may mislead a reasonable consumer into thinking they're talking to a human, unless it clearly and conspicuously says otherwise. Violations fall under the Maine Unfair Trade Practices Act.
- Utah (SB 226, effective May 7, 2025): a business using generative AI with a consumer must disclose it when the consumer clearly asks. People in licensed occupations must disclose high-risk AI interactions up front, verbally at the start of a voice interaction. An AI that discloses itself at the outset and throughout gets a safe harbor.
- California (AB 2905, 2024): automated calls using a prerecorded message must say if the message uses an artificial voice, defined as one generated or significantly altered by AI.
- California (SB 1001, 2018): bans using a bot online to mislead people about its artificial identity to drive a sale or influence a vote, unless the bot is disclosed. It is written for online interactions.
- Colorado: the 2024 Colorado AI Act was delayed, then repealed and reenacted in May 2026 as a narrower law on automated decision-making technology (SB 26-189). It requires notice when such technology materially influences consequential decisions like housing, lending or health care, not on every AI interaction.
What's changing
Many states passed chatbot laws in 2025 and 2026, but most target companion chatbots rather than customer service. Others are pending, and some would reach customer service bots directly. Bill numbers, effective dates and scope change between sessions, so check current text before relying on any summary, including this one.
Because the trend is toward more disclosure, the lowest-risk design is an AI that says it's an AI at the start of every call and answers honestly whenever asked. Building it that way once, at the platform level, is far easier than adding it state by state after a law passes.
Watch the federal side too. Check the status of the FCC's 2024 proposals on AI-generated calls before relying on this guide, since we couldn't confirm final action. The TCPA's consent rules for artificial voice calls already apply today either way.
Practical defaults for a platform
- Play a recording notice at the start of every recorded call, and keep it in the recording.
- Have the AI identify itself as an AI and name the business at the start of every call, inbound and outbound.
- Make the AI answer truthfully when a caller asks if they're talking to a person, and offer a transfer to staff.
- Get prior express consent before any outbound AI-voice call, and record when and how it was given.
- Let businesses in regulated fields, like health care or legal services, add stricter rules on top of yours.
How Wirebird handles it
Wirebird plays a consent announcement on recorded calls, configured per business, and stores it with the recording and transcript. The AI receptionist discloses that it's an AI at the start of calls where required, and platforms can set rules for every tenant, like topics the AI must never handle, with an audit log of every AI turn. Businesses choose when the AI answers and when it hands off to a person.
Sources
- 18 U.S.C. 2511 (federal Wiretap Act)
- Reporters Committee: Reporter's Recording Guide
- Washington RCW 9.73.030
- Florida Statutes 934.03
- California Penal Code 632
- Kearney v. Salomon Smith Barney, 39 Cal. 4th 95 (2006)
- FCC 24-17: Declaratory Ruling on AI-generated voices (February 2024)
- Federal Register: FCC proposed rules on AI-generated calls (FCC 24-84)
- Maine 10 M.R.S. 1500-DD
- Utah SB 226 (2025), enrolled text
- California AB 2905 (2024)
- California SB 1001 (2018)
- Colorado SB 26-189
- DLA Piper: AI disclosure laws on commercial chatbots (January 2026)