Legal
What we collect, what we never collect, and what happens when you talk to one of our AI agents — on the phone, in chat, by text or by email.
On this page
Greggy AI LLC, doing business as GreggyAI ("GreggyAI," "we," "us"), is a Virginia limited liability company — Virginia State Corporation Commission Entity ID 12023650. We build and operate AI agents that answer the phone, reply in website chat, and send and receive text messages and email on behalf of the businesses that use our platform. Our first customers are car dealerships, and the examples below use a dealership for that reason; the platform itself is not specific to one industry.
Almost everything in this policy depends on which of two relationships you have with us. Please read the one that fits you.
If you are not sure which one applies to you, ask the business you contacted — or write to us at corp@greggyai.ai and we will tell you whose records you are in and pass your request to them.
If you contact us about the platform, request early access, or are a user at a customer business, we collect:
We do not run advertising or analytics trackers on greggyai.ai. There is no ad pixel, no third-party analytics tag, no session-recording script, and no cross-site tracking. The only things we put in your browser are functional:
We use fonts served by Google Fonts on our website, which means Google receives your IP address as part of loading the font files.
This is the part most people are reading this page for. When you call, chat, or text a business that uses GreggyAI, an AI agent may handle the conversation. What follows describes exactly what happens, channel by channel. In every case the business you contacted is the one whose records these are.
Across all channels, the record we keep of a conversation is what was said in it — the transcript, and the details you chose to give, such as your name, phone number, email address, the vehicle or product you asked about, your timeline or budget if you mention it, and any appointment you book. We also keep the ordinary technical facts needed to deliver the conversation and prove what happened: which number was called or texted, when, how long it lasted, and for website chat the IP address the chat came from.
Every call to a number we answer opens with a spoken notice, before anything you say is captured. The standard wording is:
"This call may be recorded or analyzed by us or our service providers."
Three things about that notice are worth stating plainly, because they are design decisions and not accidents:
If you stay on the line after the notice, the AI agent takes the call. The standard greeting identifies it as an AI agent — for example, "Welcome to [business] — this is the AI agent." Businesses can write their own greeting, so the exact wording varies, and you can always ask the agent directly whether you are speaking to a person; it will tell you it is automated and offer to connect you with someone.
Tell the agent, and ask for a person. Where the business has given us a number to forward to, we record your phone number as having declined, forward you to a human, and route your future calls to that number straight to a person with no AI and no recording on the line. If the business has not given us a forwarding number, we cannot complete that transfer — so if this matters to you, the reliable route is to ask the business for a direct line, and to tell us so we can get the decline on file.
If nobody is available and you leave a voicemail, your message is recorded and transcribed so the business can read it and text you back. A voicemail greeting is not the AI agent and does not carry the recording notice above — the act of leaving a message is the recording.
We do not make outbound sales calls. The AI agent answers calls that come in; it does not dial out.
The chat does not start, and nothing is captured, until you agree to it. Before the chat window will accept a message you are shown a plain-language notice and have to tick a box that says "I have read and agree to the privacy policy and SMS terms above." Until that box is ticked the server refuses the conversation outright. The notice you agree to reads:
"This chat may be read by our team and internal systems to improve accuracy and service. We don't sell your information — please don't share anything sensitive you wouldn't want on file."
We keep a record that you agreed, including the wording you were shown, its version, the time, and the language it was shown in.
The chat is always started by you. It does not pop itself open, and nothing about your browsing makes it appear — see What we never do.
Your browser also keeps the conversation and your consent tick on your own device, so that you can close the window and come back to the same chat for a few hours. Clearing your browser storage clears that copy; it does not clear the business's copy.
You can ask for a human at any time using the "Talk to a sales rep" button, which pages the business's staff. A salesperson can also take over on their own initiative from their queue, without you having asked. When that happens the chat header changes to "Live — chatting with a salesperson," and their replies are shown in a different colour from the assistant's. A person who takes over can read the whole conversation up to that point, including the part you had with the AI. If the conversation goes quiet, it hands back to the AI.
If you give a mobile number and opt in, you agree to receive SMS text messages about your enquiry, appointments and related customer care from the business, sent through GreggyAI as its messaging service. Message frequency varies. Message and data rates may apply. Reply STOP to unsubscribe at any time, or HELP for help. Consent to receive messages is not a condition of any purchase.
We do not sell or share mobile information, opt-in data, or consent with third parties or affiliates for their marketing or promotional purposes. Information may be shared only with service providers that help us operate the messaging service, and only as needed to deliver it.
There are four ways a number comes to be opted in, and we record which one it was, along with the date and time:
Two honest notes on the limits of that. First, starting the conversation is treated as agreement to be answered in it: if you text a business first, the AI will reply to you in that thread without a separate opt-in, because replying to you is the point. Second, a number we hold with no record either way — for example one loaded from a business's own existing customer list — is not blocked from being contacted. What is blocked is any number that has opted out. If you never want to hear from us, the reliable step is to opt out rather than to stay silent.
Reply STOP (or STOPALL, UNSUBSCRIBE, CANCEL, END, QUIT, OPTOUT or REVOKE) to any message. The opt-out is recorded immediately and permanently against your number and you get one confirmation message. Reply HELP or INFO for help. Reply START if you later want messages again.
An opt-out is specific to the channel you sent it on. Replying STOP to a text stops texts. It does not by itself stop email from the business, and it does not remove you from its records. To stop everything, tell us or the business and ask for a full opt-out, which we can set across every channel at once.
Both sides of the conversation are stored as sent, in full, including the messages we send you. We keep the outbound copy exactly as the carrier delivered it, because that record is the proof of what was actually sent. There are no quiet hours built into the platform and no fixed cap on message frequency, which is why every disclosure says frequency varies.
Businesses can send email through the platform — appointment and enquiry follow-ups, and notifications. Before an email is queued we check whether the recipient has opted out of email, and a blocked message is stored as suppressed rather than sent, so the business can see that it was stopped. As with texts, the check is for the absence of an opt-out; it does not require a separate email opt-in.
Today there is no one-click unsubscribe link in these emails. To stop email from a business that uses us, reply to the email and say so, ask the business directly, or write to us and we will set the opt-out for you. We are building a standard unsubscribe header and link, and this section will be updated to say so when it ships rather than before.
We do not currently receive inbound email into the platform — an email you send to a business goes to that business's own mailbox, not to us.
This section is a commitment, not a description, and it is built into how the product is designed.
One caution in the other direction, because it is the honest thing to put on this page: please do not give an AI agent a card number, a bank account, a Social Security number, or a password, in chat, in a text, or out loud on a call. A conversation is stored as it happened. We strip our own system credentials out of anything the AI writes back, and keypad entries are masked as described above, but we do not detect and remove sensitive numbers from what you send us. If you need to hand over something like that, ask for a person and a proper channel for it.
For conversations handled on a business's behalf (role B), the purposes are the business's, and they are the obvious ones: answering your question, finding the product or vehicle you asked about, booking or changing an appointment, taking a message, passing you to a person, and keeping a record of the exchange so whoever picks it up next knows what was already said.
We also use conversations for two purposes of our own, and both are worth naming explicitly:
People read these conversations. The chat notice says so, and it is true of every channel. Specifically:
We do not sell personal information and we do not share it with anyone for their own marketing. We use a deliberately short list of service providers, each doing one job, each handling the information only to deliver that job for us:
We may also disclose information when the law requires it, to respond to lawful legal process, to protect our rights or someone's safety, or in connection with a merger or sale of the business — in which case this policy travels with the information.
And once more, because a messaging rule turns on it: mobile information, opt-in data and consent are not shared with third parties or affiliates for their marketing or promotional purposes, and are passed to the service providers above only as needed to deliver the messaging itself.
The agent's replies are generated by a large language model provided by Anthropic. Your conversation is sent to Anthropic to produce the reply, and — on phone calls — your speech is sent to ElevenLabs to be turned into text and to generate the voice you hear.
We do not use your conversations to train or fine-tune AI models. The platform has a switch for capturing conversations as training examples and it is turned off in our production systems. The model-training work we do is done on synthetic, machine-generated conversations that contain no real customer, and on our own separate models. We instruct our AI providers to use what we send them only to deliver the service to us.
What we do use real conversations for, as described above, is human quality review and re-running a conversation against a newer version of the AI to compare answers. Neither produces a trained model from your words, and neither sends you anything.
The agent is a language model and can be wrong. Prices, availability, specifications, appointment times and anything else it tells you are not a binding offer or a guarantee — confirm the details with the business before relying on them.
We are going to be straight with you about this rather than quote a comfortable-sounding number we do not enforce.
Conversation records do not expire on a timer today. Transcripts, call recordings, text threads, chat logs, contact records and the consent trail are kept for as long as the business we hold them for keeps its account with us, because that record is what the business relies on to serve you and what we rely on to show that a message was permitted. We delete them when the business asks us to, when a person asks us to and we are able to (see your rights), and when a business leaves the platform. Automatic expiry with a fixed retention window is something we are working on, and we will state the windows here when they are real and enforced.
Records that are not conversations expire sooner: sign-in sessions and tokens are short-lived, a website chat session identifier lasts twelve hours, and the copy of your chat kept in your own browser is good for a few hours before it is dropped.
If the business you contacted arranges financing or leasing, federal financial-privacy rules may treat what you told us — including on a recorded call — as nonpublic personal information. In that case we handle it as that business's service provider under those rules and under our contract with it, and that business's own financial-privacy notice describes your choices.
No system is perfectly secure, and we would rather you know what we have not finished than assume it. We do not currently hold a SOC 2 or ISO 27001 certification. If you need a description of our controls for a vendor review, write to corp@greggyai.ai and we will give you an accurate one.
Depending on where you live, you may have the right to confirm whether we hold personal data about you and get a copy of it, to correct it, to have it deleted, to obtain it in a portable form, and to opt out of targeted advertising, the sale of personal data, and profiling used to make significant decisions about you. We do not conduct targeted advertising, we do not sell personal data, and we do not use automated profiling to make decisions of that kind — so on this platform those three opt-outs have nothing to act on, and you can still ask us to confirm that.
How to exercise them, and what happens:
One limit you should know about. Where a conversation belongs to a business we work for (role B), we hold it on that business's behalf, and we are obliged to act on its instructions. If you ask us to delete it, we will pass your request to that business, act on it where we are able to, and tell you who to contact and what we did. We will not quietly drop the request.
Requests for access and deletion are handled by hand today rather than by a self-service button, which is why we ask for the identifiers above — it helps us find every record rather than the first one.
The platform is built for business-to-consumer communication between adults and is not directed at children. We do not knowingly collect personal information from anyone under 13, and our messaging programs are not age-gated because they are not intended for children at all. If you believe a child has given us information, write to corp@greggyai.ai and we will delete it.
We will update this page when what we do changes, and we will move the "last updated" date at the top when we do. Where a change materially affects how your conversations are handled, we will re-ask for consent in the chat rather than rely on you re-reading this page — the consent notice is versioned precisely so that bumping it prompts again.
Several sections above say plainly that something is not built yet — a fixed retention clock, a one-click email unsubscribe. Those sentences are there on purpose. When we ship them, we will change the words here, and not before.
Greggy AI LLC, doing business as GreggyAI
8401 Mayland Dr, Ste A, Richmond, VA 23294, USA
corp@greggyai.ai · 571-571-1951
For privacy requests, please put "Privacy request" in the subject line so it is routed straight away. If you are contacting us about a conversation you had with a business's AI agent, telling us the business's name and the number you used will get you an answer considerably faster.