Compliance

AI disclosure statement

Draft. This document has not been reviewed by counsel and is not in force. It is published here so that it can be read and challenged, and no school has been asked to accept it.

What the assistant is, what every caller is told about it and when, and what it will not do.

Version 0.2 · Drafted 27 October 2026 · Download as PDF

What this document is

Doras answers a school’s admissions line with an automated assistant. This statement says what that assistant is, what every caller is told about it and when, what it will not do, and what a school can and cannot change. It is written for three readers: a school’s procurement or compliance office, a regulator, and a caller who wants to know what they were talking to.

It describes the product as it is built, not as we intend to build it. Where a claim is enforced by the code rather than by a policy, it says so.

What the caller is told, and when

Every call opens with the same two sentences, in the same order, before anything else is said:

  1. That the caller is speaking with an automated assistant.
  2. That the call is recorded, or — where a school has audio retention switched

off — that it may be monitored.

The AI disclosure comes first on purpose. Somebody who is going to hang up on a machine should learn it is a machine before they are told they are being recorded.

The exact wording of the first sentence is the school’s to set. The default is “You are speaking with an automated assistant.” A school may reword it to suit its own voice, and the product refuses any wording that does not plainly identify the speaker as automated: it cannot be emptied, it cannot be reduced to something ambiguous, and there is no setting anywhere in the product that turns it off.

The second sentence is not editable at all. It is a statutory notice rather than a message, and the reasons are in the privacy policy and in the terms.

Where the caller is told it again

On request, always. If a caller asks whether they are talking to a person, the assistant says that it is an automated assistant. It is instructed never to claim to be human and never to deflect the question.

In writing, on every page a caller reaches. The reschedule and stop pages a caller opens from a text message identify the sender as a school using an automated assistant.

What the assistant will not do

  • It will not claim to be a person, under any prompting.
  • It will not give legal, medical or financial advice.
  • It will not state a job-placement rate, a tuition figure, a start date or an

accreditation beyond what the school itself has entered. Where it has not been given an answer it says so and takes a message rather than improvising one.

  • It will not continue a call with somebody who asks for a person. A caller who

asks to speak to a human is transferred to the number the school nominated, or — where the school has nominated nobody — is told so plainly and a message is taken.

  • It will not make outbound contact with anyone who has not contacted the school

first, outside 8am to 8pm where that person is, or at all once they have asked it to stop.

How a caller reaches a person

Every call can be handed to a human. The school sets the number, and a caller who asks is transferred without having to argue for it. Anything the assistant could not answer becomes an item on the school’s Queue the same minute, with the caller’s number and what they asked.

Decisions the assistant does not make

The assistant collects information and books tours. It does not decide admissions, eligibility, financial aid, or anything else about a person’s application. No consequential decision about a caller is made by this system, automated or otherwise; a school’s own staff make them from the record the assistant wrote down.

The law we build to

This is the list as of the date at the top of this document. It is not legal advice and it is not a claim that it is exhaustive.

RequirementWhere it comes fromWhat we do
Proactive disclosure that the caller is speaking with an AI, for regulated occupationsUtah SB 226 (2024), amending the Utah Artificial Intelligence Policy ActDisclosed at the start of every call, in every state, whether or not the occupation is regulated
All-party consent to recordTwelve states, and case law reaching calls placed into them from elsewhereRecording notice on every call, in every state, not removable
Consent before promotional messagingTCPA, 47 U.S.C. § 227Promotional messaging off by default; reminders and replies are transactional

What a school can change

SettingWho changes it
The wording of the AI disclosure sentenceThe school, within the rule that it must still identify the assistant as automated
Whether the disclosure is spoken at allNobody
The recording notice, and whether it is spokenNobody
The number a caller is transferred toThe school
What the assistant knows and says about the schoolThe school

Reaching us

Questions about this statement go to privacy@doras.ai, or to the Support screen in the portal, which arrives with the account already attached.