What is the difference between one-party and all-party consent?
One-party consent means a conversation can be recorded if any one participant agrees, and all-party consent means every participant has to agree.
The federal Wiretap Act follows the one-party rule, and so do most states. A smaller group of states require everyone on the call to agree. Kilpatrick's chart of US wiretap laws is one of several published lists, and the lists do not match each other. California, Florida, Illinois, Maryland, Montana, Pennsylvania and Washington appear on all of them. Other states are counted in or out depending on whether the recording is of a phone call or an in-person conversation, which is why published totals range from 11 to 14.
California is where most of the money risk sits. The California Invasion of Privacy Act (CIPA) allows $5,000 per violation or three times actual damages, whichever is greater, according to the same Kilpatrick chart. That figure applies per violation, so a business that records many calls can face a large number on paper even when no one was harmed in any ordinary sense.
A business usually cannot tell which state each participant is in. That is why many companies apply the all-party rule to every conversation. Which state's law governs a particular call is a question for counsel.
Does an AI notetaker or call-analysis vendor change who has to consent?
It can, because plaintiffs argue that a vendor which records or analyzes a conversation is an outside party listening in, separate from the business that hired it.
The older question was simple: did the people on the call know it was being recorded. With AI tools there is a second listener. A notetaker bot, a contact-center analysis service or an ordering assistant receives the conversation and may keep it or use it to improve its own product.
Holland & Knight's May 2026 review of these cases describes two lines of reasoning. One asks whether the vendor used the conversation for its own purposes. The other asks only whether the vendor was capable of doing so. The firm's view is that no bright-line rule has been established. It also warns that a standard notice that the call may be recorded may not be enough where the complaint is that a third party was analyzing the call at the same time without being announced.
For a business, the practical point is that the notice should say what is really happening. If an AI service transcribes or analyzes the conversation, say so, and say it before recording starts.
What have courts decided so far?
Plaintiffs have survived several early motions to dismiss, defendants won three notable rulings in 2026, and no notetaker case found in the research has reached a settlement, a verdict or a certified class.
The case most people watch is In re Otter.AI Privacy Litigation in the Northern District of California. On 13 August 2026 the court granted Otter's motion to dismiss in part and denied it in part. Federal Wiretap Act, CIPA and Illinois biometric claims go forward. Baird Holm's summary of the order says the court found the complaint plausibly alleged that Otter kept and used conversation data for its own commercial purposes, which let the plaintiffs argue it acted as a third-party eavesdropper. That is a ruling on the pleadings. It does not decide whether Otter did anything unlawful.
The defense wins matter just as much. In Lisota v. Heartland Dental, a federal court in Illinois dismissed a federal Wiretap Act claim over an AI call service under the ordinary-course-of-business exception, as reported by Covington. In Guerra v. Papa John's, a California federal court granted summary judgment because the statute at issue covers landline, cordless and cellular phones and the call centers used internet calling, as reported by Loeb & Loeb. In the Google contact-center AI litigation, a 22 October 2026 order dismissed one group of callers with leave to amend and sent three other groups to arbitration, as reported by OpenClassActions. That report says the court treated lack of consent as something the plaintiff has to plead.
The picture is narrow. The research behind this guide found roughly a dozen named cases. Most name the AI vendor as the defendant. The businesses sued as customers were mainly health systems and companies running contact centers. No case was found against an ordinary employer whose staff used a notetaker in meetings.
| Case | Ruling date | What happened |
|---|---|---|
| In re Otter.AI Privacy Litigation (N.D. Cal.) | 13 August 2026 | Motion to dismiss granted in part and denied in part. Wiretap Act, CIPA and Illinois biometric claims proceed. |
| Taylor v. ConverseNow (N.D. Cal.) | 11 August 2025 | Dismissal denied in a case over an AI ordering assistant. |
| Valencia v. Invoca (S.D. Cal.) | 17 November 2025 | Dismissal denied. The court used the capability approach. |
| Lisota v. Heartland Dental (N.D. Ill.) | 13 January 2026 | Federal Wiretap Act claim dismissed under the ordinary-course-of-business exception. |
| Guerra v. Papa John's (N.D. Cal.) | 23 June 2026 | Summary judgment for the defense. The statute did not reach internet-based call centers. |
| In re Google Cloud Contact Center AI (N.D. Cal.) | 22 October 2026 | One group of callers dismissed with leave to amend. Three groups sent to arbitration. |
Can a meeting tool create a voiceprint claim?
Yes, plaintiffs in Illinois argue that a tool which learns to recognize a speaker's voice has collected a biometric identifier without the written consent the state requires.
The Illinois Biometric Information Privacy Act (BIPA) covers voiceprints. The American Bar Association's litigation section describes voiceprint claims against AI tools as a new trend. It lists Cruz v. Fireflies.AI, filed in December 2025, and Basich v. Microsoft, filed on 5 February 2026 over speaker attribution in Teams. The biometric claim against Otter also survived the August 2026 ruling.
These cases are at an early stage. No ruling was found in the Fireflies case, and none of the voiceprint suits has decided whether ordinary speaker labeling in a transcript counts as a biometric identifier.
If a tool identifies speakers by voice and any participant may be in Illinois, treat speaker identification as its own consent question. Ask the vendor whether the feature can be turned off, and record whether it was on for each meeting.
What do AI medical scribes have to disclose to patients?
The duty comes from state law, and the states that have acted so far require the clinician to tell the patient before an AI tool records or documents the visit.
PYMNTS reported in November 2026 that HIPAA has no specific patient-notice rule for AI scribes, so the obligation depends on the state. Louisiana's Act 649 requires a licensed professional to verbally disclose before recording any part of a visit for AI transcription, with licensing-board discipline as the sanction. Rhode Island's S2570A requires providers who use AI to document visits to notify the patient and review the note. California's AB 3030 covers AI-generated messages about clinical information. It does not cover recordings. These descriptions come from one news report, so check the statute for your state.
Lawsuits have gone further than the statutes. Holland & Knight describes Saucedo v. Sharp HealthCare and Stewart v. Kurien, both in California Superior Court. The complaints allege that visits were recorded with ambient AI tools without consent, and that the notes the tool produced stated that patients had been told and had agreed when they had not. No rulings were reported in either case.
That second allegation is the useful lesson. A line in an AI-written note saying the patient consented is weak evidence if the tool writes that line by default. The consent record should be created separately from the scribe's own output.
What do Zoom, Teams and Google Meet now do about notetakers?
The platforms now flag or restrict AI notetakers, which helps a host control who is recording but does not produce a consent record for each participant.
UC Today reported that Microsoft introduced detection of external meeting bots in Teams in 2026. Detected bots are labeled as unverified in the lobby, the setting is on by default, and administrators can set a policy to block them for the whole organization. The same report says Zoom's AI Companion and Google Meet's note-taking feature have to be started by a host and show an indicator, and administrators can restrict them.
These controls answer whether a bot is in the meeting. They do not record who saw the indicator, when each person joined, or whether anyone objected. Baird Holm also notes that some notetakers capture audio on the person's own device without joining as a bot, which meeting-platform controls cannot see.
A banner is a notice to the room. Proof that a named participant received that notice before recording began has to be kept by the business that relies on it.
What should a per-participant consent record contain?
It should show, for each person, what they were told, when they were told, how they answered and what was recorded after that.
Lack of consent is the point these cases turn on, so the record that matters is the one that shows consent was asked for and given. Baird Holm's advice after the Otter ruling is that the questions a business will be asked are whether every participant got notice before recording started and what the vendor does with the data.
The list below is a practical checklist drawn from those questions. It is not a legal standard. Consentra's Recording Consent keeps this kind of record per participant and per jurisdiction. A spreadsheet or a field in a call system can hold the same facts if it is filled in every time.
- Who the participant was, and how they were identified (name, phone number or meeting login).
- The exact wording of the notice, including whether it said an AI service would transcribe or analyze the conversation.
- How the notice was delivered: spoken by a person, played by a phone menu, shown on screen, or sent in the invitation.
- The time the notice was given and the time recording started, so the order is clear.
- The response: agreed, declined, or continued without objecting, and how that was captured.
- What happened for anyone who joined late or declined, such as pausing the recording.
- Which vendor processed the audio, whether speaker identification was on, and whether the vendor may use the data for its own purposes.
- The state or country the participant gave, if known, and the rule the business applied.
Common questions
Which states require everyone on a call to consent to recording?
Published lists differ. California, Florida, Illinois, Maryland, Montana, Pennsylvania and Washington appear consistently, and several other states are included or left out depending on whether the recording is of a phone call or an in-person conversation. Published counts range from 11 to 14, so check the statute for any state you operate in.
Is it legal to use an AI notetaker in a meeting?
No law found in the research bans AI notetakers in ordinary business meetings. The legal questions are whether every participant consented where state law requires it, and whether the notetaker vendor needed consent of its own. Telling participants before recording starts that an AI service is transcribing, and keeping a record that you did, addresses both questions as far as a business can.
Has anyone won money from an AI notetaker company?
As of 12 November 2026, the research behind this guide found no settlement, verdict or certified class in any case against an AI notetaker or conversation-analysis vendor. The Otter case survived a motion to dismiss in part in August 2026 and is continuing.
Is saying that a call may be recorded enough?
It may not be when an outside AI service is analyzing the call. Holland & Knight warns that a standard recording notice may fall short where the complaint is that a third party was analyzing the conversation without being announced. A notice that names the AI transcription or analysis is the more careful choice, and counsel should review the wording.
Do I need separate consent for speaker identification?
In Illinois, plaintiffs argue that identifying a speaker by voice creates a voiceprint that needs written consent under the state's biometric law. Courts have not settled the point. If participants may be in Illinois, either turn the feature off or get advice on whether written consent is needed.
Does a doctor have to tell a patient about an AI scribe?
It depends on the state. PYMNTS reports that Louisiana requires a verbal disclosure before a visit is recorded for AI transcription and that Rhode Island requires notice to the patient and review of the note, while HIPAA has no specific rule on the point. California lawsuits allege recording without consent, so a practice should record consent for each visit separately from the scribe's note.
Reviewed 12 November 2026
Consentra is software. Nothing on this site is legal advice, and using Consentra does not by itself make a business compliant with any law or carrier rule. Whether a rule applies to your business is a question for your counsel.
Sources
- Kilpatrick: chart of US wiretap laws (July 2024)
- In re Otter.AI Privacy Litigation: order on motion to dismiss (CourtListener)
- Baird Holm: privacy claims against AI note-taking platform survive dismissal
- Holland & Knight: recent generative AI class actions (May 2026)
- Loeb & Loeb: lessons from Guerra v. Papa John's (July 2026)
- Covington: AI call center and the federal Wiretap Act (March 2026)
- OpenClassActions: Google contact center AI wiretap suit dismissed (October 2026)
- American Bar Association: voiceprints, AI and biometric privacy litigation
- PYMNTS: states are writing rules for AI in doctor visits (November 2026)
- UC Today: AI meeting bot controls in Microsoft, Zoom and Google
