Nevada's 2025 AI Laws and Your Phone Line

The question a law firm partner actually asked

A four-attorney personal injury firm on West Sahara called us in August. Their question was not "can AI answer our phones." It was "did Nevada make that illegal last year and nobody told us." They had read a headline about the 2025 legislative session, a compliance vendor had emailed them a deadline, and their office manager had frozen an intake project that was two weeks from going live. The same week, a used car dealer off Boulder Highway asked a version of the same thing about his after-hours line.

The short answer: Nevada did pass AI laws in 2025, they are narrow, and for most Las Vegas businesses none of them govern what happens when a machine picks up the phone. The rules that do govern it are older, and one of them is federal.

That distinction was worth two weeks of frozen project time to that firm, so here is the whole picture, bill by bill.

None of this is legal advice. It is a map of what the session produced and where the real obligations sit, so you can ask a Nevada attorney a precise question instead of a vague one.

Nevada's 2025 AI bills, in plain English

Nevada did not pass a comprehensive AI framework in 2025. It passed a handful of targeted measures aimed at specific harms, which is what The Nevada Independent's report on Nevada pursuing limited AI regulation describes. Here is the set, with the only column most owners care about on the right.

Read the right column again. Six measures, and the only one that could plausibly touch a normal business phone line is AB 406, in a narrow case. There is no general Nevada duty, as of this writing, to tell a caller they are speaking with an AI. The Nevada Independent's rundown of what the Legislature is doing to expand and regulate AI walks the same set of bills.

AB 406 is the one that gets misread

AB 406 is where the confusion starts, because it is the bill with a dollar figure attached. It was signed in June 2025 and took effect July 1, 2025. It deals with AI in education, and it restricts how certain AI systems can be represented in mental and behavioral health settings. Coverage of the 2025 session reports civil penalties up to $15,000 per violation for offering an AI that provides, or claims to provide, professional mental or behavioral health care.

For a dealership or a restaurant, that is not a live risk. For a law firm, a medical practice, or anyone running a wellness, recovery, or counseling adjacent business, it is worth ten minutes of attention, because the exposure sits in the script, not in the technology.

Here is arithmetic that makes it concrete. Assume a hypothetical intake bot at a small practice that handles 20 calls a month where the caller is in distress, and assume the script is loose enough that the bot responds with something reading as counseling rather than scheduling. If each of those were treated as a violation, 20 times $15,000 is $300,000 a month of theoretical exposure. The call volume and that characterization are our assumptions, not findings, and whether any given script crosses the line is a question for a lawyer. The point is the shape of the risk: it scales with how many conversations your script mishandles, not with whether you use AI at all.

The fix is cheap and it is editorial. An AI that says "I can get you scheduled with an attorney, and if this is an emergency please hang up and call 911" is doing intake. An AI that says "tell me more about how you have been feeling lately" has drifted somewhere it should not be. Read your prompt the way a regulator would, not the way a marketer would.

The bill that would have changed everything did not pass

SB 199 was the comprehensive one. It would have created a broad AI regulation framework with state oversight, and it failed to advance out of the Senate. For planning purposes that is the single most important fact in this post.

It matters in two directions. First, the compliance floor you build to today is set by older, general-purpose law, not by anything AI specific out of Carson City. Second, that floor is unstable in a predictable way. A comprehensive bill that dies in one session usually comes back in the next one in some form.

The practical response is not to wait. It is to build so that a future disclosure requirement is a configuration change rather than a rebuild. Concretely: your AI agent should already have a spoken disclosure line it can switch on, your call recordings should already be retrievable by date and number, and the person who owns the script should be a named human rather than a vendor support queue. We build client systems that way deliberately, which is one reason we push for on-site AI deployments in Las Vegas where the business holds its own recordings and prompt files.

If a tracker cites "SB 149," check the state

One specific piece of misinformation deserves naming, because a client got it from a paid compliance dashboard and it cost real time.

There are third-party trackers claiming Nevada has an "SB 149 AI disclosure" obligation with an October 1, 2026 deadline. Nothing in the 2025 Nevada coverage supports that. SB 149 is the Utah AI Policy Act, a different state's law. If a tracker, a vendor deck, or an inbound sales email tells you Nevada has an AI disclosure deadline next month, treat it as a sales tactic until someone shows you the bill number and the Nevada statute it amended.

That is the general test for all of this. Bill number, chamber, session, and either a signature date or a statute citation. Anything softer is marketing.

The two rules that actually reach your phone system

Here is what does apply today in Nevada when an AI answers or places a call.

The recording rule catches Las Vegas businesses first, because transcription is how AI phone agents work at all. We covered the mechanics, including the wrinkle when the caller is sitting in California, in our earlier piece on Nevada recording consent and AI phone systems.

The FCC's February 2024 position catches marketing teams instead. It ruled that AI-generated voices in robocalls are "artificial" for TCPA purposes, which means outbound AI calling needs prior express consent. Inbound is a different posture than outbound. A bot answering a call somebody chose to place to you is not the same legal object as a bot dialing a list, and if you remember one line from this post, make it that one.

Two worked examples

The four-attorney firm first. Assume 140 inbound calls a month, of which 55 arrive after hours and reach the AI. All 55 are recorded and transcribed, so all 55 need an all-party consent disclosure at the top of the call. That is 660 disclosed calls a year. Assume the disclosure line runs 4 seconds: 55 times 4 is 220 seconds a month, roughly 3.7 minutes. At an assumed 12 cents per minute of platform time, compliance costs about 44 cents a month. The firm does no outbound AI dialing, so the FCC position does not apply, and the practice area is injury rather than behavioral health, so AB 406 does not either. Total regulatory work: one sentence in the script. That is the typical case for AI at a Las Vegas law firm.

Now the 40-person dealership. Assume 900 inbound calls a month with 40 percent arriving after hours: 0.40 times 900 is 360 AI-handled calls, 4,320 a year, all needing the same disclosure. Same 4 second line, 360 times 4 is 1,440 seconds, or 24 minutes a month, about $2.88 at the same assumed rate. The inbound side is a rounding error, same as the firm.

The difference is outbound. If that dealership uses an AI voice for service reminders and unsold-lead follow-up, every number it dials needs prior express consent on file before the AI speaks, and the consent record has to be tied to the specific number dialed. That is a CRM data problem rather than a phone problem, and it is where the actual work and the actual cost live. The bigger the outbound program, the more the 2025 session recedes into irrelevance and the more the FCC position becomes the whole ballgame.

Both examples use our own assumed call volumes and an assumed per-minute rate. Run them against your real phone bill before you repeat the numbers to anyone.

Where this analysis does not hold

This post is about Nevada businesses handling Nevada callers, and it goes soft at the edges in four places.

If your callers are frequently out of state, the recording analysis changes before the Nevada analysis does, because the stricter jurisdiction tends to control. If you are in a mental health, counseling, recovery, or wellness adjacent business, AB 406 moves from background noise to the main event, and a lawyer should read your script rather than a blog post. If you run outbound campaigns at any volume, the FCC position is your governing constraint and the Nevada session barely matters to you. And if you are a healthcare provider or you handle payment card data, HIPAA and PCI obligations sit on top of everything here and are outside the scope of this piece.

One more edge: this reflects the 2025 session as reported in the sources above. Sessions change things. The next one is the next chance for this to move, and SB 199 failing is not the same as SB 199 being dead.

What to do this week

Three things, none of which require hiring anyone.

1. Pull your script and read the first 15 seconds out loud. If there is no recording disclosure, add one sentence. If any branch of your AI could be read as offering care rather than scheduling, rewrite that branch today.
2. Separate inbound from outbound on paper. Write down every place an AI voice dials out, including reminders and callbacks. For each one, name where the consent record lives. If you cannot name it, pause that campaign until you can.
3. Ask your vendor two questions in writing: where are my call recordings stored, and can I turn a disclosure line on myself without a support ticket. The answers tell you how expensive the next rule change will be.

Then confirm all of it with a Nevada attorney. This post is a map, not counsel.

If you want a second set of eyes on where your call flow sits against these rules, our free 15-minute audit will walk it with you: AI for law firms in Las Vegas.