> ## Documentation Index
> Fetch the complete documentation index at: https://docs.getnexor.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# AI disclosure and call recording rules for US agents
> US regulations on when the law requires the agent to say it is an AI or a bot, which states require everyone's consent to record a call, and how to write a compliant opener for AI voice calls, SMS and chat.
Two questions come up on every US launch: does the agent have to say it is an AI, and does it have to say the call is recorded. The safe answer to both is yes, in the first sentence. This page lists the laws behind that answer. General guidance as of September 2026, not legal advice.
## Should the agent say it is an AI?
There is no federal rule yet. The FCC proposed one in 2024 that would require an AI notice at the start of every AI call and in the consent text, and it is still pending. Several states already require it:
| State | Law | Applies to | What it requires |
| ---------- | --------------------------- | ------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------ |
| California | AB 2905 (2025) | AI voice calls | Announce at the start that the voice is artificial or AI-generated, plus the business name, address and phone |
| California | Bot Disclosure Act | Bots that sell online (web chat, Instagram, Messenger, WhatsApp) | Clear, conspicuous disclosure that it is a bot |
| Maine | LD 1727 (Sept 2025) | AI chatbots in commerce, **voice and text** | Disclose the person is not talking to a human whenever a reasonable consumer could be misled |
| Utah | AI Policy Act (2025) | Any generative AI in a consumer transaction | Answer truthfully when asked; disclosing at the start gives a safe harbor. Health, legal and accounting services must disclose proactively |
| Colorado | AI Act (Jan 2027) | AI that materially influences credit, insurance, housing, healthcare, education or employment decisions | Clear notice before the decision |
| New York | General Business Law 399-z | Every telemarketing call | Say the caller's name, the seller, and offer the seller's internal do-not-call list at the start |
| Illinois | Telephone Solicitations Act | Every solicitation | Identify, state the purpose and ask permission to continue at the start; never disguise the voice |
| Washington | RCW 80.36.390 | Every telemarketing call | Name, company and purpose within 30 seconds; hang up within 10 seconds when asked |
| Tennessee | SB 1580 (July 2026) | AI in regulated professions | The agent must never present itself as a licensed professional |
Twelve states have "companion chatbot" laws (California, Colorado, Connecticut, Georgia, Hawaii, Idaho, Iowa, Nebraska, New York, Oregon, Rhode Island, Washington). Most exclude customer-service and sales bots, but Idaho and Iowa are written broadly, so a proactive disclosure is the simplest way to stay clear of them.
Disclosing once, at the start, covers every state in this table. An agent that only admits being an AI when asked is compliant in Utah and at the federal level, but not in California or Maine.
## Does the call need a recording notice?
Federal law needs the consent of one party, which the agent provides. These states need **everyone** on the call to consent, so the notice must be spoken before anything else:
California, Delaware, Florida, Illinois, Maryland (consent must be explicit), Massachusetts, Montana, Nevada, New Hampshire, Pennsylvania, Washington.
Connecticut, Michigan, Oregon and Vermont are mixed or unsettled; treat them as all-party. Marketing consent never counts as recording consent. In Illinois, avoid any feature that identifies a person by their voice; plain transcription is fine.
## Writing the opener
Put the disclosures in the agent's opening line for every US agent. A compliant opener has four parts in this order:
1. The agent's name and your company.
2. That it is an AI assistant.
3. That the call is recorded.
4. The purpose of the call.
Example for the opening message:
```text theme={null}
Hi, this is Ana, an AI assistant calling from Stark Industries on a recorded line about the demo you requested. Am I speaking with Tony?
```
Add these rules to the agent's prompt so the behavior holds for the rest of the conversation:
* If asked whether it is a bot or an AI, confirm it. Never deny it or dodge the question.
* If the person says stop, do not call, or unsubscribe in any wording, acknowledge, end the call within ten seconds and do not contact them again on that channel.
* Never claim to be a licensed professional, a government body, or a named real person.
* For text channels, include the same identification in the first message. For SMS to US numbers, Nexor composes the first commercial message itself with the required identification and opt-out language.
## Quick answers
Only if the consent text named artificial or AI voice calls, the number, and said consent is not a condition of purchase. A generic "you may call me" consent does not cover AI voice. See [US outreach rules](/docs/en/guides/compliance/us-outreach-overview).
An AI-written text is not an "artificial voice", so federally texts are easier. State laws in Connecticut, Florida, Maryland, Oklahoma, Washington, Texas and, from October 2026, Pennsylvania require written consent for marketing texts regardless, and carrier rules require it everywhere. Nexor blocks US-to-US texts without a consent record.
If the goal is to sell or to book a sales meeting, yes. Courts look at the purpose, not whether a price was quoted.
The artificial-voice rule applies to any cell phone, including a business owner's. Treat business-to-business AI calls to mobiles as consumer calls.
They are outside the TCPA. The AI-disclosure laws above still apply (Maine explicitly, California for online sales bots, Utah when asked), and each platform has its own consent and messaging policies.