Restaurant Technology and AI

AI and Guest Privacy: Questions for Vendors

The privacy questions to ask AI phone, chat and booking vendors before guest data flows: recording consent, retention, training, texting and deletion.

AssistCall research teamPublished Reviewed by Faruque Amin, VP of Sales and Marketing7 min read

On this page
  1. Why guest privacy is a buying question, not a legal footnote
  2. Map the data before you ask a single question
  3. The questions to send every vendor
  4. Texting and calling rules to raise
  5. State privacy laws: what to ask, not what to assume
  6. Red flags in vendor answers
  7. Put the answers in the contract
  8. How to run the review in one afternoon
  9. Example: comparing two vendors
  10. Frequently asked questions
  11. References
Restaurant Technology and AI: AI and Guest Privacy: Questions for VendorsArticle coverRESTAURANT TECHNOLOGY AND AIAI and Guest Privacy:Questions for VendorsAva for RestaurantsInsightsRestaurant Technology and AI: AI and Guest Privacy: Questions for VendorsArticle coverRESTAURANT TECHNOLOGY AND AIAI and Guest Privacy:Questions for VendorsAva for Restaurants

Key takeaways

  • Half of US adults are more concerned than excited about AI in daily life[1], so guest trust is part of the buying decision.
  • Ask how calls are recorded and disclosed: federal law allows one-party consent, but some states, such as California, require all parties[2].
  • Get written answers on retention, model training, access controls, texting consent and deletion before you sign.
  • Score vendors on evidence (contracts, settings, logs), not on reassurances in a sales call.

An AI tool on your phone line or website hears and stores things guests would never write on a comment card: their mobile number, their home address for delivery, a partner's name for an anniversary booking, a child's nut allergy. Once that data leaves your reservation book and enters a vendor's system, you are responsible for choosing a vendor that handles it well.

Guests are wary. In Pew Research Center's survey, 50% of US adults said they were more concerned than excited about the increased use of AI in daily life, and only 10% were more excited than concerned[1][3]. About six in ten want more control over how AI is used in their own lives[4]. A restaurant that can explain plainly what its assistant records, and why, is in a better position than one that cannot.

How US adults feel about AI in daily lifePew Research Center survey, 2025
How US adults feel about AI in daily lifeDonut chart. More concerned than excited: 50%; Equally concerned and excited: 38%; More excited than concerned: 10%.More concerned than excited: 50%Equally concerned and excited: 38%More excited than concerned: 10%50% concerned
  • More concerned than excited 50%
  • Equally concerned and excited 38%
  • More excited than concerned 10%

Source: Pew Research Center, 2025[3]

Show the data
ItemValue
More concerned than excited50%
Equally concerned and excited38%
More excited than concerned10%

Map the data before you ask a single question

Start with a simple inventory of what the tool will touch. It keeps the vendor conversation concrete and tells you which questions matter most for your setup.

A starting inventory; add anything specific to your concept, such as event guest lists.
Data the tool handlesTypical exampleWhy it mattersAsk the vendor
Call audio and transcriptsRecording of a booking callRecording consent rules; sensitive details said aloudIs audio stored, for how long, and can recording be turned off?
Contact detailsName, mobile number, emailUsed for confirmations and texts; marketing rules applyWho can export it, and is it ever shared?
Booking notesAnniversary, high chair, wheelchair spacePersonal and sometimes health-relatedCan notes be restricted to staff who need them?
Allergy and dietary informationSevere shellfish allergyHealth information a guest expects to be handled carefullyIs it flagged, routed to a manager and kept out of marketing?
Payment informationCard guarantee for a holiday bookingPCI scope and fraud riskDoes the tool ever hear or store card numbers? It should not.
Website chat and cookiesChat history linked to a deviceState privacy laws and cookie consentWhat is tracked, and is it used for advertising?

The questions to send every vendor

Send these in writing and ask for written answers, ideally with links to the contract clauses or settings that back them up. Group them so you can compare vendors side by side.

  • Recording and disclosure: Does the assistant tell callers at the start that it is automated and whether the call is recorded? Can we edit that greeting? Federal wiretap law permits recording with one party's consent, but some states, such as California, require all parties' consent[2], so ask how the vendor handles callers from all-party states.
  • Retention: How long are audio, transcripts and contact details kept by default? Can we set shorter periods? What happens to our data if we cancel?
  • Model training: Is our guests' data used to train or improve models that serve other customers? Is that off by default, and is it stated in the contract rather than a settings page that can change?
  • Access: Which vendor staff can see our call recordings, under what circumstances, and is that access logged? Do we get role-based access for our own team?
  • Security: Is data encrypted in transit and at rest? Is there an independent security audit report you can share under NDA? How quickly will you notify us of a breach?
  • Subprocessors: Which third parties process our data (telephony, transcription, hosting)? Will you tell us before adding new ones?
  • Texting: When the assistant sends a confirmation or ordering link, how is consent captured and stored? Are texts sent from a registered sender?
  • Rights requests: If a guest asks what data we hold or asks us to delete it, can we find and delete everything tied to their number or email, including recordings?
  • Payments: Can the assistant be configured so it never asks for, hears or stores card numbers?

Texting and calling rules to raise

Many AI tools send texts: booking confirmations, waitlist updates, ordering links. Transactional messages a guest asked for are different from marketing, but the line blurs when a confirmation also promotes a weekly special. Under the Telephone Consumer Protection Act, consumers can sue for actual losses or 500 dollars per violation, whichever is greater, and courts can treble damages for willful violations[5]. FCC rules also bar telephone solicitations to residential subscribers before 8 a.m. or after 9 p.m. local time[6].

Ask the vendor whether marketing texts are separated from transactional ones, whether quiet hours are enforced by the system, and whether opt-outs such as STOP are honored automatically across every number you use.

State privacy laws: what to ask, not what to assume

Most independent restaurants will not meet the thresholds of the larger state privacy laws, but multi-unit groups might, and your vendor certainly might. California's law, for example, applies to for-profit businesses that meet at least one threshold, including annual gross revenue over 25 million dollars or buying, selling or sharing the personal information of 100,000 or more California residents or households[7]. It gives consumers rights to know, delete, correct and opt out of sale or sharing[7].

Even if you are below every threshold, ask the vendor to support the same rights for all guests. It is the simplest way to be ready if you grow, and it answers the guest who calls asking what you kept.

Red flags in vendor answers

  • We never store anything, from a tool that shows you transcripts the next day.
  • Training use is controlled by a toggle in settings, with no contractual commitment.
  • No list of subprocessors, or a refusal to notify you of changes.
  • Deletion only by support ticket with no time commitment.
  • The greeting cannot be edited to include disclosure.
  • Card numbers can be read out to the assistant and appear in transcripts.

Put the answers in the contract

A good written answer is only as durable as the agreement behind it. Before you sign, ask for the commitments that matter most to appear in the contract or a data processing addendum, not only in a help-center article that can be edited next quarter.

  • Default retention periods for audio, transcripts and contact details, and your right to shorten them.
  • A clear statement that your guests' data is not used to train models offered to other customers.
  • Breach notification within a defined number of days, with the information you will need to tell guests.
  • A current subprocessor list and advance notice of changes.
  • Return or deletion of all data within a set period after the contract ends, with written confirmation.

If a vendor will not put any of this in writing, treat that as your answer. Smaller vendors may not have polished paperwork, but they should be willing to commit to the basics in a short addendum.

How to run the review in one afternoon

  1. Fill in the data map

    Use the table above. Mark which rows apply to your setup and delete the rest.

  2. Send the question list

    Send the same list to every shortlisted vendor and ask for answers within a week, with links to contract terms.

  3. Score evidence, not tone

    Give 2 points for an answer backed by a contract clause or a setting you have seen, 1 point for a clear written answer, and 0 for vague or missing answers.

  4. Test it yourself

    Call the demo line, give a fake allergy and a fake card number, then ask the vendor to show you exactly what was stored and to delete it.

  5. Write your guest-facing line

    Draft the one or two sentences you will add to your privacy policy and greeting. If you cannot write them from the vendor's answers, the answers were not good enough.

Example: comparing two vendors

Example (fictional): a three-location taqueria group shortlisted two phone assistants. Vendor A answered every question in writing, kept transcripts for 90 days by default with a 30-day option, excluded customer data from model training in its contract, and showed that card numbers were masked in transcripts. Vendor B had a better demo voice but said training use was controlled in settings and could not say how long recordings were kept. Vendor A scored 15 of 18; Vendor B scored 7. The group chose Vendor A and added a one-line disclosure to the greeting.

Privacy is one part of the decision. Weigh it alongside the business case in measuring ROI on AI phone and chat tools, the disclosure rules in state AI disclosure laws for customer calls, and what the tool will actually do day to day, covered in AI reservation assistants: what to expect. This guide is general information, not legal advice; check specific obligations with your attorney.

Frequently asked questions

Do I have to tell callers the call is recorded?

It depends on where you and the caller are. Federal law allows one-party consent, but some states, such as California, require all parties to consent[2]. The simplest approach is to disclose recording in the greeting on every call.

Can an AI vendor use my guests' calls to train its models?

Some can, unless the contract says otherwise. Ask for a written commitment that your data is not used to train models serving other customers, and check that it is in the contract, not only in a settings page.

Should the AI assistant take credit card numbers over the phone?

No. Configure it so card details are never requested or stored, and send guests to a secure payment link or have staff handle guarantees.

Do state privacy laws apply to my single restaurant?

Often not directly, because thresholds are high, but your vendor may be covered and guests may still ask. Supporting access and deletion requests for everyone is a sensible default.

References

  1. Pew Research Center (2025). How Americans View AI and Its Impact on People and Society (opens in a new tab) ↑

  2. Legal Information Institute (Cornell Law School), US Code (2026). 18 U.S. Code § 2511 - Interception and disclosure of wire, oral, or electronic communications prohibited (opens in a new tab) ↑

  3. Pew Research Center (2025). How Americans View AI and Its Impact on People and Society (opens in a new tab) ↑

  4. Pew Research Center (2025). How Americans View AI and Its Impact on People and Society (opens in a new tab) ↑

  5. Legal Information Institute (Cornell Law School), US Code (2026). 47 U.S. Code § 227 - Restrictions on use of telephone equipment (opens in a new tab) ↑

  6. Legal Information Institute (Cornell Law School), eCFR text of FCC rule (2026). 47 CFR § 64.1200 - Delivery restrictions (opens in a new tab) ↑

  7. State of California Department of Justice, Office of the Attorney General (2026). California Consumer Privacy Act (CCPA) (opens in a new tab) ↑

Book a demo

See Ava handle your calls.

Book a short demo. We’ll walk through your calls and workflow, and agree what Ava should take care of.

Book a demo