Is Otter.ai Safe for Lawyers? What the Lawsuits and Bar Opinions Say
Otter.ai is one of the most widely used AI transcription tools in the United States. It is fast, inexpensive, and integrates cleanly with Zoom, Google Meet, and Microsoft Teams. For attorneys, though, the question is not whether Otter is convenient. It is whether feeding client conversations into a third-party cloud service is compatible with the duty of confidentiality under ABA Model Rule 1.6, whether it risks waiver of attorney-client privilege, and whether recent litigation has changed the calculus.
This article walks through what Otter's own documentation says, what the pending Brewer v. Otter.ai class action alleges, how ABA Formal Opinion 512 and NYC Bar Formal Opinion 2025-6 frame the diligence obligation, and what a defensible answer looks like for a practicing lawyer in 2026.
The Short Version
Otter.ai is a cloud service. Audio is uploaded, transcribed on Otter's servers, and stored in Otter's account infrastructure. For a lawyer, that means every recorded client meeting becomes data held by a third-party vendor, discoverable by subpoena to that vendor, and governed by that vendor's terms of service and its evolving AI-training practices.
That architecture is not, by itself, a violation of the Model Rules. But it is a set of facts a lawyer must diligence, disclose where appropriate, and be prepared to defend if the recording ever becomes relevant to a dispute. Recent bar guidance has made that diligence obligation explicit rather than implicit.
What Otter.ai Actually Does with the Recording
According to Otter's privacy policy and terms of service, when you record a meeting through Otter:
- The audio is transmitted to Otter's servers (hosted on Amazon Web Services in the United States) for processing.
- Transcripts and audio are retained in the user's Otter account until deleted.
- Otter states it uses de-identified content to train and improve its speech models, with an opt-out available on some plans.
- Otter engages subprocessors including AWS and other infrastructure providers.
- Otter will respond to lawful process, including subpoenas directed at user content.
None of this is unusual for a SaaS product. It is, however, materially different from a tool that never transmits the audio at all. For a lawyer holding privileged material, the difference matters.
The Brewer v. Otter.ai Class Action
In 2025, plaintiff Justin Brewer filed a putative class action against Otter.ai in the Northern District of California. The complaint, Brewer v. Otter.ai, Inc., alleges that Otter's Notetaker bot silently joined and recorded video meetings without the informed consent of all participants, in violation of the California Invasion of Privacy Act (CIPA) and analogous two-party consent statutes. Coverage of the filing is available from Reuters and ClassAction.org.
The suit is at an early stage and Otter denies the allegations. But the theory of the case is important for attorneys to understand even if the plaintiff never prevails: the claim is that when a cloud notetaker joins a call, every participant on that call has an independent statutory interest in whether they consented to being recorded. In a client conversation, that is not a hypothetical concern. It is the entire question of whether the recording is admissible, discoverable, or an ethics problem.
What ABA Formal Opinion 512 Requires
ABA Formal Opinion 512, issued in July 2024, is the first comprehensive ethics guidance on generative AI tools. It does not ban cloud AI. It does require that a lawyer, before using a generative AI tool with client information:
- Understand, to a reasonable degree, how the tool handles inputs, whether inputs are used to train models, and who has access to them.
- Evaluate the vendor's confidentiality, security, and data-retention practices.
- Obtain informed client consent where the tool's use presents a meaningful risk to confidentiality.
- Supervise the tool's output under Model Rules 5.1 and 5.3.
Opinion 512 is not a checklist a vendor can satisfy on your behalf. It is a diligence obligation the lawyer owes the client. Using Otter with client audio without having read its terms, without knowing where the data sits, and without any conversation with the client about it is the fact pattern the opinion is warning against.
NYC Bar Formal Opinion 2025-6
In December 2025, the New York City Bar Association issued Formal Opinion 2025-6, which builds on ABA 512 and addresses AI notetakers specifically. The opinion emphasizes three points relevant here:
- A lawyer's confidentiality duty under Rule 1.6 extends to the choice of vendor, not only to the lawyer's own conduct.
- Where an AI tool stores recordings or transcripts on vendor infrastructure, the lawyer should consider the discoverability of that data by third parties.
- Informed client consent to AI-assisted notetaking is best documented, particularly where the tool retains data or uses it for model training.
Read together, ABA 512 and NYC 2025-6 do not prohibit Otter. They do raise the floor of what a defensible workflow looks like.
Waiver and Third-Party Disclosure Risk
Attorney-client privilege protects confidential communications made for the purpose of obtaining legal advice. Disclosure to a necessary agent of the lawyer, such as a paralegal or interpreter, generally does not waive privilege. Disclosure to an unnecessary third party can.
Whether a cloud transcription vendor is a "necessary agent" is unsettled. Courts have taken different views on analogous cloud storage and e-discovery vendors. In West Technology Group v. Sundstrom, the District of Connecticut in 2024 addressed the discoverability of AI-generated meeting summaries, and the case is now frequently cited for the proposition that transcripts held by a third-party AI vendor are fair game in discovery. And in United States v. Heppner (S.D.N.Y. Feb. 10, 2026), Judge Rakoff held that a defendant's exchanges with a public generative-AI platform were not privileged: routing the content through a third-party AI provider — one whose terms the user had accepted — defeated the confidentiality element, with the court reasoning by analogy to Fourth Amendment third-party doctrine (see the Covington analysis). Heppner concerned a litigant's own AI chats rather than meeting transcripts, but its logic reaches further: content handed to a third-party AI service on the provider's terms is hard to keep inside the privilege.
The doctrinal picture is still developing. The practical picture is not: if the recording sits on a vendor's servers, the vendor can be subpoenaed, and the lawyer will have to litigate privilege from a defensive posture.
Cloud Notetaker vs. On-Device: A Side-by-Side
| Consideration | Otter.ai (cloud) | On-device notetaker |
|---|---|---|
| Where audio is processed | Uploaded to vendor servers (AWS, US) | Processed locally on the lawyer's machine |
| Third-party subprocessors | AWS and others, per Otter's list | None, if truly on-device |
| Subpoena target | Vendor can be served directly | No vendor holds the data |
| Model-training exposure | De-identified use disclosed; opt-out on some plans | No transmission, so no training input |
| Meeting-join mechanism | Notetaker bot joins the call | Captures the local audio stream; no bot |
| ABA 512 diligence burden | Vendor terms, security, retention, training | Lower surface area; still requires diligence |
| Consent posture under CIPA-style statutes | Bot presence often visible to participants | Same disclosure duty applies to the lawyer |
The table is not an argument that Otter is unusable. It is an argument that the two architectures create different risk profiles and different diligence workloads.
What a Defensible Otter Workflow Looks Like
If a firm chooses to keep using Otter, the following practices reduce, but do not eliminate, the risk:
- Sign a business associate or data processing agreement with Otter where the plan permits it, and read what it actually covers.
- Disable model-training use of your content on any plan that offers the toggle.
- Set an aggressive retention policy and delete transcripts on a schedule tied to matter closure.
- Document informed client consent to AI-assisted notetaking, ideally in the engagement letter.
- Do not record conversations subject to two-party consent statutes without explicit on-the-record consent from every participant.
- Never use a free or personal-tier account for client work.
These steps are consistent with ABA 512 and NYC 2025-6. They do not answer the underlying architectural question, which is whether the audio needed to leave the lawyer's device at all.
Questions to Ask Any AI Notetaker Vendor
- Where is the audio processed, and by whom?
- Who are your subprocessors, and can I see the list?
- Do you use my content, in any form, to train models?
- Will you sign a DPA and, where relevant, an NDA?
- What is your response protocol for subpoenas directed at customer content?
- Can I delete data such that it is unrecoverable, and on what timeline?
- Do you join meetings as a bot, and is that behavior disclosed to all participants?
A vendor that cannot answer these in writing is not a vendor a careful lawyer should be feeding client audio to.
How Basil Approaches This
Basil was built by a practicing lawyer specifically to avoid the architectural questions above. Audio capture, transcription, and summarization all run on the Apple Neural Engine on the lawyer's own Mac. Nothing is uploaded. There is no Basil server that holds client audio, because there is no Basil server in the transcription path at all. That means there is no vendor to subpoena for the recording, and no training pipeline that could ingest privileged content. Basil is privilege-safe by architecture, which reduces the number of third parties in the room rather than promising any particular legal outcome.
For in-person meetings and for Zoom, Teams, and Meet calls, Basil captures the audio locally in Computer mode without sending a bot into the call. The Basil for Law edition, launching in August 2026, adds privilege attestation, a consent log, matter organization, and Privileged & Confidential labeling on exports. Solo pricing is $19 per month with a 7-day trial. You can read more on the Basil for Law page, and related background is available in our internal write-ups on on-device AI and privilege and ABA Opinion 512 explained.
No product can eliminate waiver risk, and no product should tell you it can. What an on-device architecture does is reduce the number of third parties in the room when your client is speaking, which is the risk ABA 512 and NYC 2025-6 are asking every lawyer to think carefully about.
This article is for information only and is not legal advice.
Frequently asked questions
Is Otter.ai safe for lawyers to use with client meetings?
Otter.ai is a cloud service that uploads audio to its servers for processing and retains transcripts in user accounts. It is not per se prohibited, but ABA Formal Opinion 512 and NYC Bar Opinion 2025-6 require the lawyer to diligence the vendor's data handling, disable training use where possible, and often obtain informed client consent. Whether that risk profile is acceptable is a judgment call for the individual lawyer and firm.
What is the Brewer v. Otter.ai lawsuit about?
Brewer v. Otter.ai is a putative class action filed in the Northern District of California in 2025 alleging that Otter's Notetaker bot recorded meeting participants without their consent, in violation of the California Invasion of Privacy Act and similar two-party consent statutes. The case is at an early stage and Otter denies the allegations, but the theory highlights consent risk that attorneys should evaluate before using any bot-based notetaker.
Does using Otter.ai waive attorney-client privilege?
Disclosure to an unnecessary third party can waive privilege, and whether a cloud transcription vendor qualifies as a necessary agent of the lawyer is unsettled. Courts have taken varying positions on discoverability of AI-generated transcripts held by third-party vendors. The safer architectural answer is to keep the audio on the lawyer's own device so there is no third-party custodian to litigate around.
What does ABA Formal Opinion 512 require before using an AI tool?
Opinion 512 requires lawyers to understand how the tool handles inputs, evaluate the vendor's confidentiality and security practices, consider whether inputs are used to train models, obtain informed client consent where risk to confidentiality is meaningful, and supervise output under Model Rules 5.1 and 5.3.
How is an on-device notetaker different from Otter.ai?
An on-device notetaker processes audio and generates transcripts locally on the lawyer's machine, so the recording never reaches a vendor server. There is no third-party subprocessor, no training pipeline consuming the audio, and no vendor account that can be subpoenaed for the recording. Otter, by contrast, is architected around cloud processing and account-based storage.
Should I document client consent to AI notetaking?
Yes. NYC Bar Formal Opinion 2025-6 emphasizes that informed client consent to AI-assisted notetaking is best documented, particularly where the tool retains data or uses it for model training. Many firms now add a short AI disclosure to their engagement letters and confirm consent on the record at the start of recorded meetings.
Keep client conversations on your device
Basil transcribes and summarizes entirely on-device — no cloud, no bot, no server to subpoena. See Basil for Law →
This article is for information only and is not legal advice.