MorseSign inStart freeOpen Morse

Can my doctor use AI to record my visit? What AI scribes keep, and how to say no

The Morse teamPublished 10 min read

Yes, in most of the US your doctor can use AI to record your visit, but more and more places require them to tell you or ask you first. As of November 2026, an AI scribe counts as a recording device like any other, so states that require every party's consent to record a confidential conversation, such as California and Florida, already cover it. Rhode Island and Louisiana have added laws that make clinicians tell patients when AI documents or records a visit, Texas requires notice of AI used in care, and HIPAA lets your clinic hand the audio to a scribe company as its business associate without asking you. Wherever you live, you can say "please don't record my visit" before it starts.

On November 18, 2026 we read the scribe makers' own pages (Microsoft's Dragon Copilot, Abridge, Suki, Ambience, and Epic), HHS's guidance on business associates and de-identified data, and the state statutes on each legislature's site. Microsoft deserves credit. Its privacy white paper, updated November 12, 2026, says where your visit's audio goes, when it's anonymized, and what can't be pulled back if you change your mind, which no other vendor we read spells out. This isn't medical or legal advice.

Can your doctor record you with AI without asking?

In a one-party consent state, maybe, unless a newer law says otherwise. Federal law lets one person in a conversation agree to record it where "one of the parties to the communication has given prior consent" (18 U.S.C. § 2511(2)(d)), and most states follow that rule. California's Penal Code bars recording a "confidential communication" "without the consent of all parties" (Cal. Penal Code § 632), and Florida makes interception lawful "when all of the parties to the communication have given prior consent" (Fla. Stat. § 934.03). Our page on AI notetaker consent laws quotes all 14 states with stricter rules.

The vendors tell clinicians to ask whatever the state. Microsoft's developer guide says, "Make sure users obtain patient consent before recording the patient encounter by using Dragon Copilot" (Microsoft). Suki's documentation makes the clinician "responsible for obtaining patient consent" before any data reaches Suki (Suki). Abridge's help center gives doctors a script: "I will be using a tool that records our conversation to help me write my clinical note… Is that okay with you?" (Abridge, via the Internet Archive's October 2026 copy).

Patients have sued over visits they say were recorded without real consent. On April 8, 2026, three patients filed a proposed class action in federal court in San Francisco against Sutter Health and MemorialCare, alleging that an ambient tool "developed by Abridge AI, Inc." recorded their appointments "without patients' informed consent" (complaint). Abridge isn't a defendant, and the claims are allegations, not findings.

Which states have AI scribe laws?

For medical visits, three states have laws that make clinicians tell you about AI in your care, and none of them requires your consent.

  • Rhode Island (law since June 2026, effective on passage): providers and facilities that "employ artificial intelligence ("AI") to document in-person or telehealth visits shall notify patients of the use of AI for that sole purpose and review the AI-generated documentation for accuracy after the visit" (H 7538 Sub A).
  • Louisiana (effective August 1, 2026): a licensed professional "shall verbally disclose the use of any recording device, software, or service to a patient before recording any part of an appointment or treatment to be transcribed by artificial intelligence" (Act 649). The bill as introduced required the patient's consent (original bill). The enacted text requires only disclosure, and it shields clinicians from civil liability "in the absence of gross negligence or willful misconduct."
  • Texas (effective January 1, 2026): when an AI system is used "in relation to health care service or treatment," the provider has to disclose it to the patient "not later than the date the service or treatment is first provided" (Tex. Bus. & Com. Code § 552.051). The section ties that duty to telling people they are "interacting with an artificial intelligence system," and it doesn't name scribes.

Rhode Island's law doesn't give patients a right to refuse. The legislature's own press release says patients "have the option to opt-out if they are not comfortable" (Rhode Island General Assembly). The enacted Senate version is the same (S 2570 Sub A), and the only duties in its text are notice and review. You can still say no, but the law doesn't promise the clinic will go ahead without the tool.

California's AI disclosure law, AB 3030, covers messages to patients. It requires a disclaimer on "written or verbal patient communications pertaining to patient clinical information" made with generative AI, and it doesn't apply when a licensed provider has "read and reviewed" the message (Cal. Health & Safety Code § 1339.75, read through the Internet Archive's January 2026 copy). A scribe drafts a note for your doctor to review and sign, so the disclaimer rule doesn't fit it.

Therapy has stricter rules in places. In Maine, a licensed social worker may use AI on a recorded or transcribed session only if the client is told in writing how session data "will be stored, retained, used for training and deleted," and consents (32 M.R.S. § 7009). Our page on private AI for social workers covers the clinician's side.

What does an AI scribe record, and where does it go?

Everything said in the room, sent to the vendor's cloud. Microsoft says Dragon Copilot "securely streams Customer Data in the form of audio data (including voice of the patient, the patient's care team, and other encounter participants) to the cloud," turns it into a transcript that labels each speaker, and drafts the note from that (Microsoft privacy white paper). A spouse, a parent, or an interpreter in the room is recorded too. Ambience's page for patients says it "accurately assigns every comment to the right person," whether it's "you, a spouse, a caregiver, or an interpreter" (Ambience).

The audio can travel further than the scribe. Dragon Copilot supports add-on "AI apps and agents" from other companies, and Microsoft says they receive "patient context from the electronic health record (EHR), encounter audio turns, the clinical note, and the encounter transcript" (Microsoft). Epic's own tool, released February 4, 2026, "listens during patient visits, drafts the clinician's note, and queues up orders based on the conversation" (Epic).

How long do AI scribes keep the recording, and do they train on it?

From 30 days to whatever your clinic's contract says. Each row comes from the vendor's own page, read November 18, 2026.

ScribeHow long the audio is keptTraining on visitsSource
Microsoft Dragon Copilot (which combined DAX Copilot and Dragon Medical One in March 2025)Audio, transcript, and flowsheet data "for up to 90 days"Select audio, transcripts, and notes go to a research environment and are "anonymized within 90 days"; its models are "trained solely on anonymized data"FAQ (updated July 27, 2026); announcement; white paper (November 12, 2026)
AbridgeYour clinic's "configured retention policy," after which audio is "permanently deleted"; a clinician can delete a recording in the app "at any time," but Abridge "does not support early deletion of individual encounters outside of standard retention configuration"Its privacy policy doesn't say; an October 25, 2026 post on its speech model says "All training data has been de-identified" (licensed, public, and synthetic speech)Help center (Archive copy); tech blog
Suki (developer platform)Audio and transcript "permanently deleted after 30 days"; the note kept for the contract"de-identified and anonymized data for model training"; audio broken into chunks "such that the original audio cannot be re-constructed"Developer docs
AmbienceNot published; its privacy policy says it keeps information no longer "than needed"Patients page: used "solely to create the documentation"; privacy policy: de-identified patient information may go to third parties "when permissible" under its contractsPatients; privacy policy (December 2023)
Epic AI ChartingNot on its public pagesNot on its public pagesEpic

Every vendor that describes training says it uses de-identified or anonymized data, and that is the gap HIPAA leaves. HHS says "the Privacy Rule does not restrict the use or disclosure of de-identified health information, as it is no longer considered protected health information" (HHS). Your visit can help train a vendor's model without anyone breaking HIPAA, as long as the vendor's de-identification meets HHS's standard and its contract with your clinic allows it.

Deletion inside the app has limits too. Abridge's help center tells clinicians they can delete a recording "at any time," during a live session or after it's complete, and deleted recordings "cannot be recovered" (Abridge, Archive copy). Its article on recordings filed under the wrong patient says that once an encounter exists, it "cannot be deleted before your organization's configured retention period ends."

Does HIPAA cover the AI scribe company?

Yes, as your clinic's business associate. HHS's list of business associates includes an "independent medical transcriptionist, or the vendor of an app, that provides transcription services to a physician," and the contract between them must "describe the permitted and required uses and disclosures" of your health information and bar any others "other than as permitted or required by the BAA or as required by law" (HHS). Abridge's privacy policy, updated August 27, 2026, tells patients: "We are a 'Business Associate' to your provider and process your data solely on their instruction," and points you to your provider's Notice of Privacy Practices (Abridge).

You won't usually see that contract, so ask the clinic what it allows. Our page on HIPAA compliant AI chatbots covers what a BAA has to say, and private AI for health questions covers why HIPAA rarely reaches the chatbots you use on your own.

Can you say no to an AI scribe?

Yes, and Microsoft builds in a fallback. It says that if a patient doesn't give consent, "users can still use Dragon Copilot to record after the patient interaction is over," dictating the note themselves (Microsoft). Its white paper describes the same option "if a patient prefers not to have the encounter itself be recorded."

Say it before the recording starts, because changing your mind later gets you less. Microsoft says a clinic can file a ticket to purge "any Customer Data that can be linked back to the individual patient," but "resulting anonymized data created by Microsoft prior to the patient's revocation of consent will not be purged" (Microsoft). Before anonymization, the same white paper says that in "strictly limited scenarios" Customer Data may be reviewed by "select Microsoft employees and subprocessors," including "human annotators" who help build training sets (Microsoft).

Questions to ask at check-in or in the room:

  • Which scribe is this? The answer tells you which row of the table above applies.
  • Is the audio kept, and for how long? Abridge leaves it to your clinic's settings; Ambience doesn't publish a period.
  • Does the company use visits to improve its AI? Ask whether the clinic's contract allows de-identified use.
  • Can you dictate the note after I leave instead? Dragon Copilot is built for this.
  • If I agree, can you delete the recording before we finish? In Abridge, the clinician can do that in the app during the session.

Afterward, read the note. HHS says you have the right "to inspect, review, and receive a copy" of your records, and if something is wrong you can request an amendment, which a provider "must respond to" (HHS). Rhode Island now requires the clinician to check the AI's draft for accuracy, but nothing stops you from checking it as well.

How Morse keeps your own health questions private before and after the visit

A scribe runs on your clinic's contract, its retention setting, and its vendor's training program, and none of that is something you agreed to line by line. The questions you work through on your own, like what to ask, or what a term in the after-visit summary means, don't have to go through any of it.

  • E2EE, for the questions you wouldn't want any company to read. Draft your list for the appointment, or type in a line from your note and ask what it means. Your browser checks the secure enclave's proof that it's genuine, then encrypts the message on your device for that enclave only, and Morse's servers relay text they can't read. The enclaves are run by third-party providers, not by Morse, and this rests on the enclave hardware and attestation working as designed. E2EE runs open-weight models, has no web search or image generation, and comes with Pro at $19 a month ($17 billed yearly). How the enclave check works.
  • Private, for quick follow-ups like a lab abbreviation. New chats start here. Morse has zero data retention agreements with its inference, search, and other providers, so they've agreed to keep nothing, and where a provider confirms zero retention on each reply, Morse checks it and refuses a reply that doesn't.
  • Anonymous, to ask GPT, Claude, or Gemini to explain a diagnosis in plain words. Our proxy strips your account, email, and IP address, and the provider sees the message without knowing it's you. It may keep it for a limited time to check for abuse, under its own policy, so leave out your name, your doctor's name, and any record numbers.

Whichever mode you pick, the chat history is stored in your browser, encrypted, not on Morse's servers, and Morse doesn't train AI on your chats, sell your data, or show ads (Morse privacy). Because we don't store chats, we usually have very little to give authorities. Clearing your browser or losing the device deletes those chats, and Morse can't recover them, so copy anything you want for your next appointment into your own notes.

Bring your questions to the Morse home page without an account, or sign up free and move to Pro when you want E2EE.

Common questions

Will saying no to the AI scribe change my care?

The state laws we read don't say either way. Microsoft tells clinicians that if a patient doesn't consent, they can still use Dragon Copilot to dictate the note after the visit.

Can people at the scribe company hear my visit?

Some can, before the recording is anonymized. Microsoft says select employees and subprocessors may review identifiable data in 'strictly limited scenarios,' including human annotators who help build the anonymized training sets.

I said yes and changed my mind. Can I get the recording deleted?

Ask the clinic, because the vendor works for it. Microsoft says the clinic can file a ticket to purge data that links back to you, but anonymized data made before you revoked consent stays.

Ask what you actually want to ask.

80+ models, characters, and images, with every chat encrypted and kept on your device. Free to start.

Start a private chat

Read next