Safety and privacy
What is done, stated plainly
Written for a practice owner, and for the insurer or practice manager reading over their shoulder. It says what is done, not how it feels.
What is stored
The plan you attach to a link: items, stages, timing, and fees if you included them. Your note to the patient. The conversation between the patient and the assistant, after de-identification. The flags raised on each conversation. Practice settings, the pages you have edited, and who signed in.
If you choose to, the patient themselves: their name, date of birth, your own patient number for them, and an email address if you want Chairside to send them their link. That is held encrypted, it is never sent to the AI service, it is never put in an email to you, and none of Chairside's own admin screens show it; every time a record is opened, the log says who opened it. The assistant does not know your patient's name — their page greets them by first name, written by the page itself.
What is stored sits in a database in the Oceania region. The AI service that writes the answers runs overseas, and receives the conversation and the plan but never the patient's identity. We will name every service we use, in writing, on request.
What is removed
Before a conversation goes to the AI service, and again before it is stored, names, dates of birth, phone numbers, email addresses and street addresses are removed. Appointment dates are kept, because they matter to the plan. The patient is asked not to type personal details in the first place.
Who can read what
A practice sees conversations for its own patients and nobody else's. Sign-in needs a username, a password and a code from an authenticator app. Accounts are created from a set-up link emailed to the person’s own address. Roles exist for dentists, hygienists, practice managers and front desk, and one login can belong to more than one practice.
A practice can choose whether de-identified conversations are shared with Chairside to improve the tool. When sharing is on, the patient's page and the assistant both say so. When it is off, they are not shared.
What the patient is told
That nobody reads the chat as it happens. That it is not a way to contact the clinic and cannot book, reschedule or pass on a message. That their dentist may read the conversation later to improve the tool. That they should not type personal details. These are stated on the page, not buried in terms.
What happens in an emergency
The assistant is not monitored, and it says so. If a patient describes pain, swelling, bleeding, trauma or distress, the reply points them to the clinic, an after-hours service, the emergency department, or 000 and Lifeline, early in the reply and before any other content.
The conversation is flagged. Nobody reads the chat as it happens, and no one at Chairside reads it at all. What the practice can choose is to be emailed when something is flagged as an emergency or as urgent — a count and a link, never a word of what the patient wrote. The assistant itself never claims a message has been passed on or promises that anyone will reply, because a chat is not a way to get help. The patient is told all of this before they start: that nobody is reading, that the clinic may be told a conversation needs a look, and that it could be hours or days before anyone does.
What the assistant will not say
It answers from your practice's published patient information and the patient's plan, and from nothing else. It does not diagnose. It does not propose treatment that is not in the plan. It does not quote or estimate a cost that is not written in the plan. It does not draft messages or emails for the patient to send. When it has nothing on a topic, it says so and the topic is recorded as one your pages do not cover.
How content is controlled
Every page the assistant answers from is editable by the practice, one paragraph at a time. A practice's wording outranks the shipped wording. When Chairside changes a default page, the practice's edits stay and the practice is told what changed.
How long links last
You choose how long a link stays open when you make it: 90 days, 180, or a year. Beyond that, a link closes after 90 days without use and everything on it is removed 90 days after it closes. A practice can close or reissue a link at any time. Five wrong codes lock a link.
Deleting things
Deleting a link deletes what was asked on it. Deleting a practice takes its settings, its pages, its links and its conversations with it. The nightly clear-out records what it removed each time it runs, so deletion can be shown to have happened rather than asserted.
What is flagged
Five things, and the practice decides which of them reach an inbox: an emergency (harm, facial swelling, bleeding that will not stop, trauma, someone unsafe at home), something urgent (severe pain, a problem since treatment), a caution (very anxious, using something to cope, asking after drugs, talk of complaints), ready to book, and keen. Emergency and urgent can email you; a caution reaches you the evening before you see that patient.
The policy itself
This page is what is done, in plain words. The privacy policy is the same thing in the shape an insurer or a practice manager expects, and it names every service that touches the information.
Questions
If your insurer or practice manager has a question this page does not answer, ask it when you ask for early access and we will answer it in writing.