Nevada Recording Consent Rules for AI Phone Agents

Does Nevada law let an AI phone agent record calls?

Not without consent from every party on the line. NRS 200.620 makes it unlawful to intercept a wire communication unless two conditions are both satisfied: the interception is made with the prior consent of one of the parties, and an emergency situation exists. In Lane v. Allstate Insurance Co., 114 Nev. 1176 (1998), the Nevada Supreme Court read those clauses as conjunctive, which leaves consent from everyone on the call as the only ordinary lawful route. If your AI agent records, it discloses and gets agreement before capture starts.

That is the short answer. What follows is the part almost nobody in this market writes down: what the statute actually says, exactly where the ambiguity sits, what a compliant script sounds like, and what it costs to be wrong.

Where does the all-party rule come from if the statute says one party?

Read the operative sentence. NRS 200.620(1) bars interception of a wire communication unless (a) it is made with the prior consent of one of the parties, and (b) an emergency situation exists and it is impractical to get a court order first. Standing alone, clause (a) looks like a one-party rule. The word joining (a) and (b) is "and," so one-party consent by itself never clears the bar outside an emergency.

Why did the Nevada Supreme Court read it that way?

The reasoning in Lane was structural, not textual gymnastics. NRS 200.650 covers in-person conversations and expressly allows monitoring when "authorized to do so by one of the persons engaging in the conversation." The Legislature put a one-party authorization in the in-person statute and left it out of the telephone statute. Lane treated that omission as deliberate. Justice Maupin dissented, arguing you cannot intercept your own conversation.

Worth knowing about the case: Lane was a former Allstate employee who secretly recorded more than 700 calls with company officials to build his employment claim. The court was not blessing corporate recording practice. It was construing a criminal statute against a private party, and the construction it landed on binds you too.

Is transcribing a call different from recording it?

This is the question every AI voice vendor gets asked and almost none answer honestly. Nevada law does not resolve it. NRS 200.610 defines "person," "wire communication," and "radio communication." It does not define "intercept," the verb the entire prohibition turns on. I could not find a Nevada appellate decision applying it to real-time speech-to-text.

So the honest position is that a system which transcribes live and stores no audio sits in undefined territory, and undefined territory is not the same as safe. The federal wiretap definition points the other way: 18 U.S.C. 2510(4) defines intercept as the "aural or other acquisition of the contents" of a communication, and "or other acquisition" comfortably covers a transcript. Federal law is one-party consent, so it is not your binding constraint in Nevada, but it shows that drawing a line at stored audio is not a line the drafters of wiretap statutes have historically drawn.

Practical rule: if the words of the call leave the call and land anywhere durable, treat it as recording and get consent. That includes vendor-side logs, quality-review queues, analytics dashboards, and anything routed into model training. If you are running a custom AI assistant on a third-party API, ask the vendor in writing what is retained, for how long, and where.

Does the front desk follow the same rule as the phone line?

No, and this trips people up. In-person conversations run under NRS 200.650, which permits recording when one participant authorizes it. Telephone calls run under NRS 200.620 as construed by Lane, which requires everyone. Same customer, same building, two different consent standards depending on whether they walked in or dialed in.

What does federal law require when the AI places the call?

Recording consent and calling consent are separate problems, and clearing one does not clear the other. On February 8, 2024 the FCC adopted a Declaratory Ruling, FCC 24-17 in CG Docket No. 23-362, holding that AI-generated voices are "artificial" under the Telephone Consumer Protection Act. Outbound calls using them require the called party's prior express consent, and prior express written consent for telemarketing.

The exposure there is per call, not per campaign. Under 47 U.S.C. 227(b)(3), a private plaintiff can recover $500 per violating call, trebled to $1,500 if the violation was willful or knowing. Multiply that by a dialer list and the arithmetic gets ugly fast.

Does the TCPA apply when the customer calls you?

No. The artificial-voice restriction attaches to calls that are made or initiated by the caller, not to the technology answering an inbound line, as the FCC's own framing made clear when it adopted the ruling (Wiley analysis, February 2024). An AI receptionist picking up your published number is outside that rule. It is inside NRS 200.620 the second it records.

That asymmetry matters when you are scoping a build. Inbound answering and missed call recovery carry a Nevada recording question and no federal consent question. Outbound follow-up carries both.

What does a compliant disclosure actually sound like?

Short, first turn, before capture, and genuinely answerable. Something in this shape:

Hi, this is the automated assistant for Desert Auto Group. This call is recorded so we can book and confirm your appointment. Is that all right?

Three things make it work. It identifies the business. It states recording as a fact before any audio is written. It asks a question the caller can answer no to. A disclosure the caller cannot decline is notice, not consent, and Lane's reasoning does not obviously convert continued conversation into agreement. Build a no-recording path, or a transfer to a human, and test it the way you would test any other branch.

What are the penalties for getting it wrong?

Under NRS 200.690, willfully and knowingly violating NRS 200.620 through 200.650 is a category D felony, which NRS 193.130 sets at 1 to 4 years in state prison and a fine up to $5,000. The civil side is the one that usually reaches a business: actual damages or liquidated damages of $100 per day of violation with a $1,000 floor, whichever is greater, plus punitive damages and reasonable attorney fees.

Read that civil clause as an operator, not a lawyer. Per-day liquidated damages with a statutory floor and fee shifting is a structure that makes small claims economically worth bringing, and makes a systematic recording default across thousands of callers worth aggregating.

Does it matter where your vendor's servers sit?

It can, and there is a case directly on point. In Ditech Financial LLC v. Buckles, 133 Nev. Adv. Op. 64, 401 P.3d 215 (2017), the Nevada Supreme Court answered a certified question from federal court and held that NRS 200.620 does not reach recordings of interstate calls when the act of recording occurs outside Nevada. Ditech's call centers recorded in Arizona and Minnesota, so those states' laws governed.

Do not turn that into a compliance strategy. Buckles involved an out-of-state servicer calling into Nevada. A Las Vegas business recording its own local callers, with a Las Vegas phone number and a Las Vegas customer base, is not that fact pattern. The safer read is that Ditech tells you where a court will look, not that offshoring your audio pipeline buys you an exemption. If keeping recordings inside your own walls is the goal, that is an on-premise deployment question, not a jurisdiction-shopping one.

How do you decide what your agent has to say?

1. Determine call direction. Inbound only, outbound only, or both. Outbound pulls in the TCPA layer on top of Nevada law.
2. Inventory persistence. Audio files, transcripts, vendor logs, analytics stores, training pipelines. If the content of the call lands anywhere durable, treat it as recording.
3. If recording, put an affirmative consent prompt in the opening turn, before the first frame is written.
4. Build and test the decline path: a no-recording mode, or an immediate human transfer.
5. For outbound, confirm you hold prior express consent for the specific number, written consent if the call is telemarketing, and that you logged the source and timestamp of that consent.
6. Store the consent event on the call record itself, not in a separate system nobody can join against it later.
7. Have a Nevada attorney review the script and the retention policy before the first live call.

Which rule applies to which situation?

What is still genuinely unsettled?

Three things, and any page telling you otherwise is overselling. Whether real-time transcription is an "intercept" under a statute that never defines the term. Whether continuing a call after a disclosure constitutes consent under Lane's reasoning. And what the FCC ultimately requires of AI voice: its proposal to define AI-generated calls and mandate a disclosure at the start of each one, released September 2024, remains a proposal, and the Eleventh Circuit vacated the related one-to-one consent rule on January 24, 2025.

Sources also conflict on the basic characterization. Several recording-law references still file Nevada under one-party consent while their own text describes a hybrid rule, one-party for in-person and all-party for phone. The hybrid description is the accurate one. Separately, Nevada's 2025 session produced AI laws covering mental health services, school counseling, and political advertising, but no general requirement that a business disclose a customer is speaking to an AI. Absence of a disclosure mandate is not a reason to skip disclosure, and it says nothing about the recording rules above.

This post is not legal advice, and reading it does not create an attorney-client relationship. It is my own reading of Nevada statutes and published decisions as of July 2026, written by an engineer, not a lawyer. Consult a licensed Nevada attorney before deploying any system that records, transcribes, or places calls.