Policies

AI assistant notice

Version
1.0
Status
Draft, pending legal review
Effective
Upon publication
Last updated

Status: Draft, pending legal review. Version 1.0. This page will show an effective date and a “What changed” note once it is final. Our disclaimers page explains what a draft means.

What this notice covers

This notice explains the help assistant on amynpos.com. It says what the help assistant is, what it can and cannot do, what happens to your questions, and what you should not type into it.

The website and account terms define the words we use in every policy. In this notice, “we” and “Amyn” mean Amyn, the company building the Amyn point of sale. “You” means anyone who uses the help assistant. “The help assistant” means the automated assistant on the website. Anthropic provides it (model: Claude). “The provider” means Anthropic.

This notice works alongside our privacy notice, website and account terms, cookie notice and acceptable use policy, and with the merchant agreement once you have an account. For anything about the help assistant, this notice has the detail. If two policies disagree, the order set out in the website and account terms applies. If we add other AI features later, we will update this notice before they go live.

Amyn and the help assistant are not directed at anyone under 18.

What the help assistant is

The help assistant is an automated program. It is not a person. It answers questions about Amyn using the pages we have published on this website, such as our features, our approach to pricing and our policies.

The help assistant cannot retrieve anything from an Amyn account. It cannot see sales, customers, staff, payouts or sign-in details. It cannot change anything. It only answers questions. It does receive whatever you type, so do not enter card, customer or staff information. See “What not to enter” below.

Who provides the model

The help assistant runs on Claude, a language model made by Anthropic. Amyn writes the instructions the help assistant follows and supplies the knowledge base it draws on. Anthropic supplies the model that turns your question into an answer.

We may change which Claude version we use. If we ever switch to a different provider, we will update this notice before the change takes effect.

What the help assistant can and cannot do

The help assistant can:

The help assistant cannot:

The help assistant may be wrong. It can misread a question, mix up details or give an answer that is out of date. When it does not know something, it is designed to say so and to point you to a person. If an answer differs from a published page or policy, the published page is correct.

Please treat an answer as a starting point. For anything that matters to your business, check the published page or ask us directly. Do not use an answer to decide about anyone’s job, housing, credit, insurance or health. Those decisions need a qualified person.

Where the help assistant runs

The chat window is part of this website, which Netlify, Inc. hosts. When you send a message, it goes first, exactly as you typed it, to a small server program that we run with Supabase, Inc. That program then runs the card-number check described below, passes the chat to the provider, and stores the log. Netlify, Supabase and Anthropic are the only companies that handle your chat. Our privacy notice says what we have confirmed so far about where each of them processes data.

The help assistant sets no cookie and stores nothing in your browser. When you leave the page, the chat on your screen is gone. Our cookie notice lists everything this website stores in your browser.

What is sent to the provider

When you ask a question, we send the following to the provider:

Before anything is sent, we run a check that removes text that looks like a card number. Our screening is imperfect. It catches obvious patterns only, and anything it misses goes to the provider and into our log as you typed it. That is why we ask you not to enter card, customer or staff information, or any other sensitive detail, at all.

We do not send your name, email address, IP address or account details. The help assistant does not know who you are unless you type it.

What we keep

We keep a log of each chat for 30 days, then a scheduled job deletes it. Database backup copies follow the backup cycle we will publish before launch, and the provider’s own copies follow the retention setting described below. A log holds the messages, with the card-number patterns our check caught removed, and the time of each message. It also holds a scrambled code for the chat session and a count of how much of our monthly budget the chat used. We keep logs so we can:

To enforce the limits described below, we also keep a scrambled form of your IP address for a short window. We do not keep the address itself for that purpose. Like every web service, our hosts keep short-lived server logs that include IP addresses and browser types. Our privacy notice covers those logs.

Only a small number of people at Amyn can read chat logs, and only for these reasons. We do not use chat logs to train any AI model, and we do not share them with advertisers or data brokers. We share them only as our privacy notice describes, for example when the law requires it.

The provider handles what we send under its own terms for business customers. We will use the shortest retention setting the provider offers for our account. Before the help assistant goes live, we will confirm that those terms do not let the provider train its models on our chats. We will link to the provider’s current terms, and state the retention setting, on this page before launch.

You can ask us to delete a chat. See “How to contact us” below and tell us when the chat took place and what it was about. When we can identify the chat from what you tell us, and have verified the request where we need to, we delete it from our active logs. If we cannot identify it, we will tell you, and the chat expires at the end of the normal log period. Copies in our database backups and at the provider follow the retention schedule we will publish on this page before launch. We plan to give each chat a private reference you can quote, so we can find it with certainty.

What not to enter

Please do not type any of the following into the help assistant:

The help assistant does not need any of this to answer a question about Amyn. If you entered something sensitive by mistake, contact us as soon as you can and we will delete that chat as “What we keep” describes. Until then, what you typed stays in our log and with the provider.

Limits on use

The help assistant limits how many questions one visitor can ask in a short period. It also has a monthly cap on total use, which we set. When the cap is reached, the help assistant pauses until the next month.

When the help assistant is paused or unavailable, you will see a message saying so and a way to reach a person. We may pause the help assistant, change how it works, or remove it. If we remove it for good, we will say so on this page. We may also stop serving a visitor who misuses it.

How you will know it is an AI

The help assistant is always labelled as an AI assistant. It will not claim to be a person, and it will tell you it is automated if you ask. A way to contact a human is shown next to it at all times.

The help assistant inside the dashboard

If we add the help assistant to the dashboard, the signed-in area where a merchant runs its account, it will work the same way. It will answer from our published help content only. It will not see your sales, customers, staff, payments, settings or any other account data.

If we ever propose an assistant that can see account data, we will treat that as a material change. We will publish a new version of this notice and email account holders at least 30 days before it takes effect.

How you may use the help assistant

Our acceptable use policy applies to the help assistant. In short, you must not:

You own what you type. You may use the answers you receive. We claim no rights in your questions or the answers beyond what we need to run the service and keep the logs described above.

If an answer causes a problem

The help assistant is a free help tool. It can be wrong, and its answers are not legal, tax, financial or other professional advice. We cannot promise its answers are complete, correct or current, so please check the page an answer points to before you act on it. What we are and are not liable for, including for the help assistant, is set out in “Limits on what we owe you” in our website and account terms. Counsel: liability for reliance on assistant answers, and whether the website-terms cap works for visitors who pay no fees, to be settled with counsel.

If you use the help assistant to break the law, someone may bring a claim against us because of it. In that case we may ask you to cover the costs that your unlawful use caused. Counsel: scope of this indemnity for unlawful use of the help assistant to be settled with counsel.

How to contact us

Email info@amynpos.com or use the contact form on the website. Put “AI assistant” in the subject line so we can route it. If you need this document in another format, ask us; our accessibility statement explains how we help. We will publish our mailing address here before launch. Every policy, with its version and effective date, is listed at /policies.

Use the same channels to ask us to delete a chat, to report a wrong answer, or to ask anything about this notice.

Changes to this document

Every version of this document carries a version number, an effective date and a “What changed” note. Earlier versions stay in an archive linked from /policies. If we make a material change, we email account holders at least 30 days before it takes effect. Smaller changes, such as fixing a typo or a link, take effect when posted. This is version 1.0, and no earlier version exists.

A material change to this notice is one such as a new provider, a longer log period or any access to account data.

All policies