When a dealership plugs an AI agent into its lead channels, customer conversations start flowing through third-party software: names, phone numbers, what documents a buyer holds, roughly what they earn, what they can put down. That's useful data — it's also regulated data, and dealers are held to a higher standard than most retailers realize. This isn't legal advice, but it is the practical map every dealer should have before signing an AI contract.
Why dealerships aren't ordinary retailers here
A store that arranges financing isn't just selling cars — under federal law it's acting as a financial institution. That triggers two frameworks dealers already live with, whether they know it or not:
- The Gramm-Leach-Bliley Act (GLBA) governs how nonpublic personal financial information is handled and shared, and requires privacy notices to consumers.
- The FTC Safeguards Rule requires an actual information-security program: a designated qualified individual, risk assessments, access controls, encryption, and — critically for this topic — oversight of service providers who touch customer data. Your AI vendor is a service provider. The rule expects you to vet them and bind them contractually to protect the data.
On top of that sit state privacy laws — California's CCPA/CPRA is the strictest, and states like Colorado, Texas and Virginia have their own — which give consumers rights to know, delete and opt out of the sale of their data. And if your AI or BDC sends texts or makes calls, the TCPA consent rules apply exactly as they would to a human.
What data an AI actually touches
Be concrete about the inventory: contact details; full conversation transcripts; qualification answers (document type such as ITIN or passport, income method, down payment); appointment history; sometimes photos, voice notes and CRM records the AI writes to. Individually mundane — together, a detailed financial profile of a consumer, often one from a community that is understandably sensitive about where documents like ITINs get mentioned.
What can be used — and what shouldn't be
Legitimate use: processing the conversation to serve that customer — answering, qualifying, booking, logging to your CRM. That's the service the consumer is knowingly participating in.
Gray-to-red zone:
- Training shared AI models on your customers' conversations. Some platforms use client data to improve models that serve all their clients. That may be acceptable with proper anonymization and contract terms — or a serious problem. It must be disclosed and controllable.
- Collecting more than the job needs. There is no reason for a chat agent to take Social Security numbers, full bank details or document photos in a Messenger thread. Qualification needs categories ("has ITIN"), not credentials. If an AI is configured to ask for sensitive identifiers in chat, that's a design failure.
- Selling or sharing lead data. Conversations your customers had with "your" assistant ending up enriching someone else's marketing database is both a legal exposure and a trust bomb.
Questions to put to any AI vendor — in writing
- Where is conversation data stored, and is it encrypted in transit and at rest?
- Is our data used to train models that serve other customers? Can we opt out?
- What is the retention period, and what happens to the data when we cancel?
- Who are your subprocessors (cloud providers, model providers), and what are they allowed to do with the data?
- Will you sign terms obligating you to safeguard the data, consistent with our Safeguards Rule program?
- How do you handle a consumer deletion request under state privacy law?
- Can we export and review every transcript?
A serious vendor answers these quickly. Evasion on the training question or the subprocessor list is a signal in itself.
Practical policies that keep you clean
- Minimize by design: configure the AI to collect qualification categories, never identifiers. Move document collection to the showroom or a secure portal.
- Disclose the bot: honesty aside, California's bot-disclosure law requires it in sales contexts, and customers respond fine to a well-built assistant that doesn't pretend.
- Honor opt-outs everywhere: a "stop" in any channel ends messaging in that channel, logged in the CRM.
- Fold the vendor into your Safeguards program: list them in your risk assessment, keep the signed data terms on file, review annually.
- Limit internal access: transcripts contain financial details; not everyone with a login needs them.
The bottom line: AI doesn't change your privacy obligations — it concentrates them. The same data that once lived in a salesperson's memory and a paper folder now sits in a vendor's database, which makes your obligations more auditable, not less real. Dealers who treat the AI vendor like any other financial service provider — vetted, contracted, monitored — get the benefits without inheriting someone else's data practices as their own liability.
Want to see this working on your own inventory? UCallNow builds AI sales agents, BDC teams, Facebook Marketplace auto-posting and dealer websites for dealerships across the United States — in English and Spanish. Try SOPHIA live or see every solution and price.