DealerSpark

The Dealer's AI Fast-Track

The Dealer's AI Fast-Track › Part 6 of 11

Part 6

Risk, compliance, and a staff policy you can copy today

None of this requires a law degree, but it does require house rules before your staff starts pasting things into chatbots — because they already are, whether you've sanctioned it or not.

The AI use already happening in your store

Before the rules, an uncomfortable inventory. Walk any dealership today and some of this is going on, on personal phones and free accounts, with the best of intentions and zero oversight:

None of this makes your people bad actors — it makes them early adopters with no rules. The wrong response is a ban, which just drives it further underground. The right response is the rest of this part: clear rules, paid accounts with data protections for the sanctioned uses, and a one-page policy that tells everyone where the lines are. You can't govern what you pretend isn't happening.

About that self-serve sales training: a rep asking ChatGPT for closing advice is showing you exactly the hunger for coaching your store isn't feeding. But what a generic assistant returns is the averaged advice of every sales book ever written — with none of your process, your inventory, your market, or your word tracks in it. It isn't wrong; it's generic, and it's identical to what the rep across town is getting from the same prompt. The fix isn't to stop them — it's to give the practice structure: the role-play prompt in Part 2 grounded in your store's context, and when you want every rep practicing daily with scoring a manager can see, the platform version in Part 3.

The regulatory picture in six paragraphs

The FTC's position is simple and dealers should tattoo it somewhere: there is no AI exemption from consumer-protection law. A deceptive price claim, a fake review, an invented offer — the FTC treats an AI-generated one exactly like a human-written one, and it has brought enforcement actions against companies both for deceptive claims made by AI tools and for deceptive claims about what their AI could do. "The chatbot said it, not us" is not a defense. Your tools' claims are your claims.

The Safeguards Rule already covers you. Dealerships that arrange financing are financial institutions under GLBA, and customer financial data is exactly what must not leak into unsecured tools. An employee pasting a credit application into a free chatbot is a data-handling incident, full stop. Your written information security program should name AI tools explicitly.

Your state may have its own privacy law. Around twenty states now have a comprehensive consumer privacy law, and the details differ in ways that matter to a dealer: how many consumers' data you must handle before the law applies, whether you get a grace period to fix a violation, the penalty per violation, and whether the financial-institution exemption covers your whole business or only the GLBA-covered data. The trend is toward tighter. Montana is a useful example: its law took effect in October 2024, and amendments effective October 2025 cut the threshold in half (to 25,000 consumers), ended the cure period, added civil penalties of up to $7,500 per violation, and narrowed the exemption to GLBA-covered data. Translation for a dealer in a state like that: your credit and deal data is still GLBA territory, but marketing lists, website tracking, and AI chat logs likely are not — and more stores, especially groups, clear the threshold than did a year ago. Whether your state has a law, and where you sit under it, is a thirty-minute conversation with your attorney or your state association's counsel (see Part 8).

Calls and texts have their own tripwire. The FCC has confirmed that AI-generated voices in outbound calls fall under the TCPA's robocall rules — consent requirements apply. Any AI tool that calls or texts customers outbound needs its consent story checked before you turn it on, not after. The opt-out side moved in 2026 too: the FCC rewrote its consent-revocation rules, effective November 25, 2026, changing how a customer's "stop" must be honored across your calling and texting. A separate FCC proposal to require AI disclosure at the start of a call is still pending, not final. You don't need to master the details — you need every vendor that calls or texts for you to explain, in writing, how they capture consent and honor opt-outs under the current rules.

Your OEM has opinions too. Franchise agreements, brand standards, and co-op advertising rules all apply to AI-generated content exactly as they do to human-made content — and OEMs are starting to issue AI-specific guidance. Loop in your OEM rep before deploying anything customer-facing that carries the brand.

Who writes the AI rules is still being fought over. A December 2025 executive order pushed for one national AI framework and directed federal agencies to challenge state AI laws; Congress has debated overriding those laws but, as of this update, has not passed anything that does. So state laws remain in force and the picture will keep shifting. Don't wait for it to settle: the four rules below hold up under any outcome, because they're built on things that aren't in dispute — protect customer data, don't deceive, disclose, verify.

The four rules that prevent 90% of problems

  1. No customer PII in public AI tools. Names, SSNs, credit apps, deal jackets, DMS exports — never into a consumer chatbot. Paid business accounts with data protections, and vendor tools under contract, are the right home for anything sensitive.
  2. A human reviews anything customer-facing. AI drafts the ad, the email, the reply; a named person approves it. Especially anything with a price, a payment, or an availability claim in it.
  3. Disclose AI where a customer might reasonably think it's a person — chat widgets, voice agents, texting tools.
  4. Verify before you rely. AI states wrong things confidently — invented statistics, misremembered regulations, wrong incentive details. Anything going into a contract, a disclosure, or a compliance decision gets checked against the source.

What can go where: the traffic-light table

DataPublic chatbot (free account)Paid business AI accountContracted vendor tool
Marketing copy, inventory descriptions, public info Fine Fine Fine
Internal docs: schedules, meeting notes, aggregated numbers Avoid Fine Fine
Customer names + contact info Never Only if your data terms cover it If contracted for it
Credit apps, SSNs, deal jackets, DMS exports Never Never Only under a security-reviewed contract

The one-page staff policy — copy this verbatim

Fill in the brackets, have your attorney spend thirty minutes on it, and put it in front of every employee. Done is better than perfect: the risk isn't an imperfect policy, it's the six employees already using free chatbots with no rules at all.

[Dealership Name] — AI Use Policy

  1. Approved tools. Employees may use the following AI tools for work: [list — e.g., the dealership's paid ChatGPT/Claude business accounts and contracted vendor tools]. Other AI tools may not be used for dealership work without approval from [name/role]. Dealership email, calendars, files, CRM, and DMS may not be connected to any personal AI account, AI browser, or automation tool.
  2. Customer information stays out. Never enter customer names, contact information, Social Security numbers, credit applications, deal documents, or DMS data into any AI tool that is not an approved, contracted vendor system. When in doubt, leave it out.
  3. Humans approve customer-facing content. Anything a customer will see or hear — ads, emails, texts, website content, review replies, quotes — must be reviewed and approved by [role] before it goes out, regardless of who or what drafted it. No AI tool may communicate a price, payment, or availability claim to a customer without human review.
  4. We disclose AI. Where customers interact with an AI system (chat, phone, text), the system must identify itself as automated.
  5. Verify before relying. AI output used for any business decision, disclosure, or compliance matter must be checked against the original source. AI is a drafting assistant, not an authority.
  6. Report problems immediately. If an AI tool says something wrong to a customer, or you realize sensitive data went into the wrong tool, tell [name/role] the same day. No blame for reporting — only for hiding.
  7. Questions. If you're unsure whether a use is okay, ask [name/role] before doing it.

I have read and understood this policy.    Name ______________________    Date ____________

By Chad Nethery, Co-founder, DealerSpark.Ai · A free guide for dealers · Updated November 2026

The Dealer's AI Fast-Track: all parts