A New AI Disclosure Law Took Effect August 2. The One That Actually Reaches Your Store Is Seven Years Old.

A compliance date landed on August 2. Most dealers can safely ignore that one.

On August 2, 2026, the first wave of the California AI Transparency Act became operative. The date was chosen, not stumbled into: when the legislature amended the act in October 2025, it moved the original January 1, 2026 start to August 2 specifically to line up with the EU AI Act's transparency timeline (Bradley Arant Boult Cummings, via National Law Review, July 2026). Two of the largest AI rulebooks in the world switched on the same morning.

If you run a store and caught the headlines, the reasonable question is whether your AI phone agent now has to announce itself. Under this particular law, for almost every U.S. dealership, the answer is no. Understanding exactly why is more useful than the answer itself, because the reason points straight at the rules that do apply - and those are older, quieter, and easier to trip over.

Myth: The California AI law that took effect August 2 means my AI phone agent has to disclose that it is AI.

Data: The act's obligations fall on "covered providers" - companies that create a generative AI system with more than one million monthly visitors or users that is publicly accessible in California - and by the statute's own terms they cover AI-generated image, video, and audio content, not text. A dealership is not a covered provider. But California's bot-disclosure statute has been on the books since 2019, Utah's disclosure rule since 2024, and Section 5 of the FTC Act the entire time. Those reach retailers.

Sources: Morgan Lewis, August 2026; Bradley / National Law Review, July 2026.

What actually became operative, and for whom

The California AI Transparency Act was created by SB 942 in September 2024 and substantially rewritten by AB 853 in October 2025. Its first tranche of duties applies to a narrow class of companies: persons who create, code, or otherwise produce a generative AI system with more than one million monthly visitors or users that is publicly accessible in California.

Source: Morgan Lewis, "New California AI Disclosure Rules Become Operative," August 3, 2026.

Two limits matter for auto retail. First, the act does not apply to AI-generated text at all. Second, penalties run at $5,000 per violation with each day of noncompliance treated as a separate violation, enforced by the attorney general - and there is no private right of action, so this is not a plaintiff's-bar exposure. Later phases reach generative AI hosting platforms and large online platforms on January 1, 2027, and camera and recorder manufacturers on January 1, 2028. A dealership does not sit in any of those buckets either.

Where it can still reach your store - through the contract, not the counter

There is one thread worth pulling. An AI voice agent produces audio, and the large model vendors underneath most voice products clear a million monthly users without breaking a sweat. The flow-down provision means a covered provider that licenses its system has to contractually require licensees to maintain latent disclosure capability, and has to cut the license inside 96 hours if it learns that capability was disabled.

Whether the act's audio provisions were drafted with real-time telephony in mind - as opposed to synthetic media files - is exactly the kind of question reasonable lawyers will read differently, and it is unresolved. That ambiguity is the point. It makes this a vendor-contract review item rather than an operations change: ask your AI vendor, in writing, which entity in the stack is the covered provider and what the license terms now require of you. If you already run the kind of vendor security and data diligence the last two years have made routine, this is one more line on the same checklist.

The rules that do reach a dealership

Here is the practical map, which looks nothing like the headlines.

Two details in that table deserve a manager's attention.

The California bot law applies to online communications - the statute reaches communications appearing on a public-facing website, web application, or digital application. That wording puts your website chat widget squarely inside the rule and leaves a live inbound phone call in genuinely murkier territory. The counterintuitive result: the AI on your website is more clearly regulated than the AI on your phone. Most stores assume the reverse.

Utah's rule is the one most likely to surface in an actual conversation. As amended by SB 226, it requires disclosure when a consumer makes a clear and unambiguous request to know whether they are dealing with a human or with AI, with proactive disclosure required in higher-risk interactions - including when a business is giving advice on financial matters. A caller asking whether they are talking to a real person is not a hypothetical. It is a Tuesday. Your agent should have a scripted, honest answer and a fast path to a person, which is what a well-designed human handoff requires anyway.

Colorado is a lesson in not writing policy from headlines

If you built a compliance plan around AI-law coverage in early 2026, you have already rebuilt it twice.

Colorado's original AI Act, SB 24-205, was the country's most sweeping state framework: high-risk system classification, mandatory risk management programs, annual impact assessments, a duty of care. Its compliance date moved from February 1, 2026 to June 30, 2026. Then, on May 14, 2026, Governor Polis signed SB 26-189, which repealed and replaced it outright. The high-risk classification, the risk management programs, the annual impact assessments, and the duty of care are gone. What remains is a narrower transparency statute covering automated decision-making technology used in consequential decisions, effective January 1, 2027, with attorney general rulemaking due by that same date.

Meanwhile the federal posture moved in the other direction. On July 1, 2026, the FTC published a proposed policy statement arguing that AI companies distorting model outputs for undisclosed ends may be deceiving consumers under Section 5, and asserting that Colorado's AI Act "appears to coerce companies into altering the output of their AI models" and is "impliedly preempted to the extent it conflicts with a federal regulatory scheme." Public comments closed July 31, 2026.

The takeaway is not that regulation is toothless. It is that the map is still being drawn, and any AI policy a store adopts should be written to survive the next amendment - which means anchoring it to conduct you would defend regardless, not to one statute's checklist.

Disclosure is a customer-experience decision before it is a legal one

Strip out the statutes and a plainer question remains: what happens to your close rate if a shopper finds out after the fact that the friendly voice was software?

Cars.com surveyed U.S. car shoppers November 4 through 10, 2025, and found a market well past squeamish about AI in general. Forty-four percent had already used AI tools to shop for a vehicle. Among AI users, 97% said it would influence their purchase decision and 73% said it saved them time. Seventy-one percent expressed at least a moderate amount of confidence in AI to provide unbiased, accurate information.

And 63% were concerned that AI tools might recommend cars in a biased way.

That last number is the one to plan around. Shoppers are not afraid of AI; they are alert to being steered by it. The same survey found 64% open to additional recommendations from a salesperson on the lot while still preferring to handle budget and financing directly with the dealership. Trust in the tool and suspicion of the motive coexist comfortably inside the same customer - which is why disclosure costs a store almost nothing and concealment can cost a deal, a review, and the referrals behind it.

None of this is legal advice, and the state-by-state picture is moving fast enough that your own counsel's read beats a blog post's. But the customer-experience math does not depend on which statute wins.

The market context that makes this worth ten minutes

This is not a year with slack in it. Cox Automotive's updated read puts July 2026 new-vehicle sales at roughly 1.36 million units and a 16.3 million SAAR - below its own 16.7 million forecast, with sales down 1.8% year over year and retail specifically down 1.7% (Cox Automotive, updated August 4, 2026). Year to date the market is running at a 16.0 million pace against a 15.8 million full-year forecast.

In a market that is not adding customers, avoidable own-goals cost more, and getting caught not saying it was a bot is an avoidable own-goal. The upside case for AI on the phone has always been about not losing the calls you already paid to generate. A disclosure fumble turns that asset into a liability for free.

Five things worth doing this month

• Ask your vendor which entity is the covered provider. Get it in writing, along with what the August 2 flow-down terms now require of you as a licensee.
• Pick one disclosure posture and apply it everywhere. Website chat is your clearest legal exposure; the phone is your clearest trust exposure. Splitting the policy by channel creates exactly the inconsistency an enforcer or a reviewer notices first.
• Script the direct question. When a customer asks whether they are speaking with a person, the agent should say so plainly and offer a human immediately. That single line satisfies the Utah-style trigger and, more to the point, satisfies the customer.
• Keep disclosure and consent in separate boxes. Saying this is an AI assistant does not give you permission to call or text anyone. Outbound still lives under the TCPA and state rules - see our rundown on AI calling and compliance.
• Log the interactions. If a question ever arrives, the record is the defense. It is also how you find out whether AI-handled calls are actually converting, which matters more day to day than any of the above.

The measured takeaway

August 2 was a real deadline for a small number of very large AI companies. It was not a deadline for car dealers. What it did do is make visible a question most stores had never formally answered: when a customer talks to your AI, do you tell them, and can you prove what was said?

The stores that answer that deliberately - once, in writing, across the phone and the website - will spend the next two years adjusting a policy. The stores that never answered it will spend those years reacting to whichever headline lands next. Given that the rules changed three times in Colorado alone inside eighteen months, the deliberate version is cheaper.

If you are evaluating an AI voice agent and want to pressure-test how it handles disclosure, human handoff, and clean CRM logging against your own call volume, that is the conversation we have every day at Carbuki. We build AI voice agents for U.S. dealerships, and we are glad to help you think it through.

Sources

• Morgan Lewis, "New California AI Disclosure Rules Become Operative," August 3, 2026: https://www.morganlewis.com/pubs/2026/08/new-california-ai-disclosure-rules-become-operative
• Bradley Arant Boult Cummings, via National Law Review, "California's Ongoing AI Regulation: Key Deadlines Arriving in 2026 and Beyond," July 9, 2026: https://natlawreview.com/article/californias-ongoing-ai-regulation-key-deadlines-arriving-2026-and-beyond
• California SB 942, California AI Transparency Act (bill text): https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202320240SB942
• California SB 1001, Bolstering Online Transparency (B.O.T.) Act (bill text): https://leginfo.legislature.ca.gov/faces/billTextClient.xhtml?bill_id=201720180SB1001
• Perkins Coie, "I Am Robot: California's New Law Requires Disclosure of Use of Bots": https://perkinscoie.com/insights/update/i-am-robot-californias-new-law-requires-disclosure-use-bots
• Davis Wright Tremaine, "Utah Enacts Multiple Laws Amending and Expanding the State's Regulation of the Deployment and Use of Artificial Intelligence," April 2025: https://www.dwt.com/blogs/artificial-intelligence-law-advisor/2025/04/utah-regulation-ai-policy-mental-health-chatbots
• Davis Wright Tremaine, "Colorado AI Act Repealed and Replaced by Narrower Statute Focused on Transparency Requirements and Enhanced Consumer Rights," May 2026: https://www.dwt.com/blogs/privacy--security-law-blog/2026/05/colorado-ai-act-repeal-new-transparency-law
• Colorado General Assembly, SB 26-189, Automated Decision-Making Technology: https://leg.colorado.gov/bills/sb26-189
• Federal Trade Commission, "FTC Seeks Public Comment on Policy Statement Addressing AI Accuracy," July 1, 2026: https://www.ftc.gov/news-events/news/press-releases/2026/07/ftc-seeks-public-comment-policy-statement-addressing-ai-accuracy
• Cars.com "AI in Car Shopping Consumer Survey" (fielded November 4-10, 2025), reported by Digital Dealer, December 18, 2025: https://digitaldealer.com/news/ais-influence-on-car-shopping-skyrockets-new-survey-reveals/168732/
• Cox Automotive, "New-Vehicle Sales Pace Strengthens in July as Buyers Shrug Off Economic Headwinds" (updated August 4, 2026): https://www.coxautoinc.com/insights/cox-automotive-forecast-july-2026-us-auto-sales-forecast/