If you run AI voice or chat outreach into the United States, the compliance question you keep getting asked — “do we legally have to say it’s an AI?” — has no single federal answer. The FCC proposed one in August 2024 and has not adopted it. So the operative disclosure rules are state rules, and they are a genuine patchwork: different triggers, different wording, different timing, different enforcers.

Most of what circulates about that patchwork is wrong. Texas SB 140 in particular is described across vendor blogs as an “AI disclosure law” with commencement dates ranging from September 2024 to September 2025. We went to the enrolled text at the Texas Legislature. SB 140 is not an AI law at all. Below is the map rebuilt from statute text only — and where we could not confirm a rule at a legislature, statute or attorney-general primary, we left the state out rather than guess.

Zian AI’s outbound appointment setters and support agents are built to run phone, SMS, email and WhatsApp outreach, so this patchwork is a design constraint for us, not a footnote. Our position is that the disclosure line belongs in the opening seconds of every conversation regardless of state, because the cost of saying it is a few words and the cost of omitting it is a deceptive-trade-practice claim. If you want to see how that is built into an agent rather than bolted on, Apply For Partnership.

The federal layer: consent rules, not disclosure rules

Two federal things are already settled and they are frequently confused with disclosure.

First, consent. On 8 February 2024 the FCC released a Declaratory Ruling (FCC 24-17, CG Docket No. 23-362, adopted 2 February 2024) recognising that calls made with AI-generated voices are “artificial” under the Telephone Consumer Protection Act. The Commission’s own release says the ruling “takes effect immediately”. The practical consequence is that a synthetic-voice outbound call inherits the TCPA’s existing artificial-or-prerecorded-voice regime: prior express consent, and prior express written consent where the call is telemarketing, under 47 CFR §64.1200(a).

Second, identification. This is the underrated one. 47 CFR §64.1200(b)(1) requires that all artificial or prerecorded voice telephone messages must, “At the beginning of the message, state clearly the identity of the business, individual, or other entity that is responsible for initiating the call” — and if a business, the registered name. Paragraph (b)(2) requires a callback number, and (b)(3) requires an automated opt-out mechanism within two seconds of that identification for telemarketing to residential lines. So there is already a federal first-breath duty on AI-voice calls. It is a duty to identify the caller, not to admit the caller is a machine.

Third, the gap. The FCC’s Notice of Proposed Rulemaking (FCC 24-84) was published in the Federal Register on 10 September 2024. It proposes a definition of “AI-generated call” and proposes “requiring callers using AI-generated voice to, at the beginning of each call, clearly disclose to the called party that the call is using AI-generated technology”. Comments closed on 10 October 2024 (reply comments 25 October 2024). Searching the Federal Register by docket on 18 August 2026, CG Docket No. 23-362 returns exactly one document — that 2024 proposed rule. There is no final rule. We covered the proposal in detail in our breakdown of what NPRM 24-84 would mean for AI voice vendors; nothing has changed its status since.

State disclosure duties sit on top of the federal consent duties, not instead of them. Announcing that you are an AI does not cure a missing consent, and having consent does not excuse a missing state disclosure.

The verified state map

Every row below was read at the statute, enrolled bill or attorney-general primary linked in the last column. Checked 18 August 2026.

State	Statute + section	What triggers the duty	What must be disclosed, and when	Enforcement hook	Verified at primary
Utah	Utah Code §13-77-103, §13-77-104, §13-77-105 (eff. 7 May 2025)	A “supplier” using generative AI — expressly including audio — to interact with an individual in a consumer transaction	That the individual “is interacting with generative artificial intelligence and not a human”, if they make a “clear and unambiguous request”. Regulated occupations must disclose proactively in high-risk interactions “verbally at the start of a verbal interaction”	Division of Consumer Protection; violation of §13-11-4(1); administrative fine up to $2,500 per violation; court disgorgement; $5,000 per order violation	le.utah.gov §13-77-103
Maine	10 M.R.S. §1500-DD (P.L. 2025, c. 294; reallocated from §1500-Y)	Using an AI chatbot — defined to include “aural communications”, so voice counts — in trade and commerce in a way that “may mislead or deceive a reasonable consumer” into thinking they are talking to a human	Notice “in a clear and conspicuous manner that the consumer is not engaging with a human being”. No timing is specified, so it must land before the misleading impression forms	Violation of the Maine Unfair Trade Practices Act — Attorney General action plus private remedies under 5 M.R.S. §213	Maine Revisor §1500-DD
California (BOT Act)	Bus. & Prof. Code §§17940–17943 (SB 1001; operative 1 July 2019)	Using a bot to communicate online with intent to mislead about its artificial identity and to knowingly deceive about the content of the communication, in order to incentivise a purchase or sale of goods or services, or to influence a vote in an election. “Online” is defined as websites, web applications and digital applications — phone calls are outside it	Disclosure “clear, conspicuous, and reasonably designed to inform persons with whom the bot communicates or interacts that it is a bot”. Disclosure is a complete defence	No private right of action created by the chapter; exposure runs through California’s unfair competition law	leginfo §17941
California (SB 243)	Bus. & Prof. Code §§22601–22606 (Stats. 2025, c. 677; eff. 1 Jan 2026)	Operating a “companion chatbot” where a reasonable person would be misled into believing they are interacting with a human. Customer-service and business-operations bots are expressly excluded	“A clear and conspicuous notification indicating that the companion chatbot is artificially generated and not human”	Private right of action: injunctive relief plus the greater of actual damages or $1,000 per violation, plus fees (§22606)	leginfo ch. 22.6
New York	General Business Law art. 47, §§1700–1704	Operating an “AI companion” that sustains a human-like relationship. Systems used solely for customer service are excluded from the definition in §1700	Notification “at the beginning of any AI companion interaction… and at least every three hours for continuing AI companion interactions”, verbally or in writing, that the user is not communicating with a human	Attorney General enforcement under §1703; no private right of action	NY Senate GBL §1702
Colorado	HB 26-1263 (signed 29 May 2026); requirements apply from 1 January 2027	Operating a “conversational artificial intelligence service”: an AI system “accessible to the general public” that “primarily simulates human conversation and interaction through adaptive textual, visual, or aural communications”. Services “primarily designed to provide commerce-related or transactional assistance”, expressly including “product or service recommendations” and “customer service”, are carved out of the definition (§6-1-1701(3.5)(b)(II))	Operators must “clearly and conspicuously disclose to a user that a conversational artificial intelligence service is artificial intelligence” — at the beginning of the user’s first interaction each day, at least every three hours in a continuous interaction (or as a persistent disclosure), and in response to user prompts. Plus age estimation, minor protections and a self-harm protocol	Attorney General — operators report annually to the AG; Colorado Department of Law rulemaking underway	Colorado GA HB 26-1263
Texas	Bus. & Com. Code ch. 552 (TRAIGA, HB 149; eff. 1 Jan 2026)	A governmental agency making an AI system available to interact with consumers; separately, a health care provider using AI in treatment	“Before or at the time of interaction, that the consumer is interacting with an artificial intelligence system”. No equivalent duty on private sellers	Attorney General has exclusive authority (§552.101); the chapter expressly creates no private right of action	capitol.texas.gov HB 149

What Texas SB 140 actually does

SB 140 (89th Legislature, Regular Session) was signed on 20 June 2025 and, per the Texas Legislature Online history page, is “Effective on 9/1/25” — September 2025, not 2024. Its caption is “Relating to certain definitions relating to the regulation of and private rights of action arising from certain solicitation-related communications.”

Reading the enrolled text: it amends Business & Commerce Code §§302.001, 302.003 and 304.005 and adds to chapters 304 and 305. It redefines “telephone solicitation” as a “call or other transmission, including a transmission of a text or graphic message or of an image”, which is how SMS and MMS marketing get pulled into the Texas telemarketing registration and do-not-call regime. And it makes violations actionable as deceptive trade practices under Chapter 17, Subchapter E of the DTPA, with the associated private remedies.

There is no mention of artificial intelligence, automated voices or AI-generated content anywhere in the bill. So the scope answer is “voice and SMS”, and the date answer is 1 September 2025 — but the AI answer is “this statute says nothing about AI”. What SB 140 does do is raise the stakes on every other outbound rule in Texas, because a DTPA private right of action is a far more expensive failure mode than a regulator letter. For AI callers, the exposure is registration, do-not-call and misrepresentation — not a missing robot announcement.

Where the map has holes

We planned to include Illinois and Florida and dropped both: we could not locate a general AI call-disclosure duty at a statutory primary in either. Treat that as “not found”, not “cleared” — proving a negative across fifty statute books is not something a blog post should claim to have done.

Colorado deserves a second flag. SB 26-189 was signed on 14 May 2026 and repeals and re-enacts the 2024 Colorado AI Act, replacing the algorithmic-discrimination framework with an automated decision-making technology regime whose consumer notice duties apply from 1 January 2027. That notice is about a technology’s role in a consequential decision, not about announcing a bot — a distinction we unpack in our piece on what SB 26-189 means for sales teams. Colorado’s actual bot-announcement duty lives in HB 26-1263, and the Colorado Department of Law has rulemaking on both statutes in progress, so the operational detail is not final yet.

Two traps worth naming, because both bit us during verification. Utah’s disclosure sections were enacted as Chapter 13-75 in SB 226 and then technically renumbered to Chapter 13-77 to avoid duplication with other 2025 bills, and the older §13-2-12 was repealed outright — anything citing §13-2-12 today is stale. Maine’s provision was enacted as 10 M.R.S. §1500-Y and reallocated to §1500-DD. If your compliance register cites either original number, it is pointing at nothing.

What to actually implement

You cannot run seven scripts. Script once to the strictest standard on the map and deploy it everywhere.

• Disclose in the first breath, unconditionally. The strictest live rule is the Utah safe harbour: clear and conspicuous disclosure at the outset of the interaction and throughout it. Meeting that meets Maine, meets New York’s opening notification, and pre-complies with Colorado from January 2027. It also happens to be what the FCC has proposed federally.
• Stack it onto the §64.1200(b)(1) identity line you already owe. One opening beat: named entity, the fact that it is an AI assistant, then reason for calling. Our disclosure scripts that do not kill the conversation covers the wording that survives contact with a real prospect.
• Answer the direct question truthfully and instantly. Utah’s duty is reactive — it fires on a “clear and unambiguous request”. Your agent must never deflect “am I talking to a robot?”, and that behaviour needs a regression test, not a prompt instruction.
• Log the disclosure as an event, not a transcript line. Store a timestamped, per-call record that the disclosure was delivered, with the script version. If enforcement runs through a consumer-protection statute, your evidence problem is proving delivery across thousands of calls, not proving the words existed.
• Keep consent and disclosure in separate columns. They are separate duties with separate defences. Our outreach compliance overview maps how TCPA consent, the Spam Act and GDPR interact for multi-region teams.
• Do not confuse disclosure with call authentication. Announcing your AI does nothing for delivery or answer rates if your calls are being flagged upstream; that is a STIR/SHAKEN and caller-reputation problem, covered in our caller-ID authentication guide.

This is a fast-moving area. Everything above was verified at primary sources on 18 August 2026; Colorado’s rules are mid-rulemaking and at least four of these statutes have been amended or renumbered within eighteen months of enactment. Re-check before you rely on it. If you would rather the disclosure logic be a platform property than a checklist your reps carry, Apply For Partnership.

Sources and who owns each figure

Figure / claim	Owner (organisation)	Where it’s published	Date checked
AI-generated voices are “artificial” under the TCPA; ruling takes effect immediately (FCC 24-17, adopted 2 Feb 2024, released 8 Feb 2024, CG Docket 23-362)	Federal Communications Commission	docs.fcc.gov DOC-400393A1	18 Aug 2026
47 CFR §64.1200(b)(1)–(3): identity at the beginning of the message, callback number, 2-second opt-out	Office of the Federal Register / FCC (eCFR)	ecfr.gov §64.1200	18 Aug 2026
NPRM 24-84 proposes AI disclosure “at the beginning of each call” and an “AI-generated call” definition; comments closed 10 Oct 2024; no final rule in the docket	Federal Communications Commission, published by the Office of the Federal Register	federalregister.gov 2024-19028	18 Aug 2026
Utah §§13-77-103 to 13-77-105: reactive disclosure, safe harbour, $2,500 per violation	Utah State Legislature (Office of Legislative Research and General Counsel)	le.utah.gov §13-77-104	18 Aug 2026
Maine 10 M.R.S. §1500-DD, incl. “aural communications” and reallocation from §1500-Y	Maine Office of the Revisor of Statutes	legislature.maine.gov §1500-DD	18 Aug 2026
California Bus. & Prof. Code §17941 and the “online” definition in §17940	California Office of Legislative Counsel	leginfo.legislature.ca.gov §17941	18 Aug 2026
California Bus. & Prof. Code §§22601–22606 (SB 243), customer-service exclusion, $1,000 per violation	California Office of Legislative Counsel	leginfo.legislature.ca.gov ch. 22.6	18 Aug 2026
New York GBL §1702 three-hour notification rule and §1700 customer-service exclusion	New York State Senate (official law database)	nysenate.gov GBL §1702	18 Aug 2026
Colorado HB 26-1263 definition, AI disclosure duty and 1 Jan 2027 application date	Colorado General Assembly	leg.colorado.gov HB 26-1263	18 Aug 2026
Colorado SB 26-189 repeals and re-enacts SB 24-205; ADMT point-of-interaction notice	Colorado General Assembly	leg.colorado.gov SB 26-189	18 Aug 2026
Colorado ADMT and chatbot rulemaking in progress	Colorado Department of Law / Attorney General	coag.gov/ai	18 Aug 2026
Texas SB 140 effective 1 Sep 2025; caption; redefines “telephone solicitation”; DTPA hook; no AI provisions	Texas Legislature (Texas Legislative Council)	capitol.texas.gov SB 140 enrolled	18 Aug 2026
Texas HB 149 (TRAIGA) §552.051 governmental-agency disclosure; §552.101 AG exclusivity; eff. 1 Jan 2026	Texas Legislature (Texas Legislative Council)	capitol.texas.gov HB 149 enrolled	18 Aug 2026

Frequently asked questions

Does Texas SB 140 require me to disclose that my caller is an AI?

No. The enrolled text of SB 140 contains no reference to artificial intelligence. It broadens the definition of “telephone solicitation” to include text, graphic and image transmissions, amends Business & Commerce Code chapters 302, 304 and 305, makes violations actionable as deceptive trade practices, and took effect on 1 September 2025. Texas’s AI disclosure duty is in TRAIGA (HB 149, Bus. & Com. Code §552.051) and applies to governmental agencies and health care providers, not private sellers.

Is there any US federal law requiring an AI to announce itself on a call?

Not yet. What exists is a consent regime plus an identification regime. The FCC’s rules at 47 CFR §64.1200(b)(1) require every artificial or prerecorded voice message to state clearly, at the beginning of the message, the identity of the business responsible for initiating the call. A duty to announce that the voice is AI-generated is only proposed, in NPRM 24-84, which remains a proposed rule as at 18 August 2026.

Does California’s BOT Act cover phone calls?

No. Bus. & Prof. Code §17940 defines a bot as an “automated online account” and defines “online” by reference to public-facing websites, web applications and digital applications. Voice calls sit outside the chapter. California’s companion-chatbot law (SB 243) is also unlikely to catch ordinary sales outreach, because §22601 expressly excludes bots used for customer service and business operations.

Which single state rule should I design my script to?

Utah. Its safe harbour at §13-77-104 rewards disclosing at the outset of the interaction and throughout it, and its underlying duty fires whenever a person clearly asks whether they are talking to a machine. A script that satisfies both also satisfies Maine’s misleading-impression test and New York’s opening notification, and pre-complies with Colorado’s January 2027 requirement.

Do these state laws apply to B2B calls, or only consumers?

It varies and you should check per state. Utah’s duty attaches to a “consumer transaction” as defined in §13-11-3, Maine’s to “trade and commerce”, and Colorado’s HB 26-1263 to services “accessible to the general public”. None of these framings map cleanly onto a B2B-only carve-out, so assuming your outbound is exempt because you sell to businesses is a bad default.

What happens if a state disclosure duty conflicts with the federal rules?

In practice they stack rather than conflict, because the federal rules govern consent and caller identity while the state rules govern whether you must say the caller is a machine. Utah makes the layering explicit: §13-77-106 states that the chapter “does not displace any other remedy or right authorized under… the laws of this state; or federal law.”