Otter.ai Alternative for Law Firms: What to Look For (and What to Avoid)
Otter.ai is one of the most widely recognized AI meeting notetakers on the market. It transcribes calls, summarizes meetings, and integrates with the video platforms most knowledge workers already use. For a marketing team or a product manager, that is often enough. For a practicing attorney, it usually is not.
Legal work introduces a set of constraints that consumer and general-business notetakers were never designed around: the attorney-client privilege, the duty of confidentiality under ABA Model Rule 1.6, the work-product doctrine, client-consent obligations, and the growing scrutiny of third-party AI vendors after ABA Formal Opinion 512 (July 2024) and NYC Bar Formal Opinion 2025-6 (December 2025).
This article walks through what a law firm should actually evaluate when replacing Otter.ai — the architectural questions, the ethics questions, and the practical questions — and where the common alternatives fall short.
Why Attorneys Are Rethinking Otter.ai
Otter.ai is a cloud service. Audio is streamed from the meeting to Otter's servers, transcribed there, stored there, and processed there. That architecture is not a defect of the product — it is how most SaaS notetakers work — but it creates several issues that matter more in a legal context than in a general business one.
The most visible recent flashpoint is Brewer v. Otter.ai (N.D. Cal. 2025), a putative class action alleging that Otter's Notetaker bot recorded conversations without the consent of all participants and that transcripts were used to train the company's models. Whatever the eventual merits, the case highlights the exposure a firm inherits when it routes client conversations through a third-party recording and transcription service.
Layer on top of that US v. Heppner (S.D.N.Y. Feb 2026), in which Judge Rakoff held that a litigant's chats with a public AI platform were not privileged, drawing an analogy to the third-party doctrine. The court's reasoning — that disclosing content to a third-party service can defeat confidentiality expectations — is directly relevant to any cloud notetaker that ingests attorney-client conversations.
The Six Questions Every Firm Should Ask a Notetaker Vendor
Before signing (or renewing) with any AI notetaker, ask the vendor to answer these in writing. If they cannot, that is your answer.
- Where does the audio go? On-device only, or uploaded to a server?
- Who is a subprocessor? OpenAI, Anthropic, AWS, Google, Deepgram, AssemblyAI — who touches the data?
- Is my content used for training? Ever. Under any tier. With what opt-out.
- Will you sign a DPA and an NDA? On the standard plan, not just enterprise.
- What is the retention model? Can transcripts be deleted, and is deletion actually deletion or just soft-delete?
- How does a bot join calls? Does a visible participant join, and does the platform (Zoom, Teams, Meet) notify or consent-prompt attendees?
These questions map directly to the risk analysis ABA Opinion 512 asks lawyers to perform before adopting a generative AI tool: understand the technology, evaluate confidentiality risk, and obtain informed consent where appropriate.
What to Avoid in an Otter.ai Alternative
1. "Zero data retention" as the whole story
Several vendors advertise "zero retention" or "no training on your data." Both can be true and still leave the transcript flowing through a third-party server in real time. Retention policy is a promise; architecture is a fact. A subpoena to the vendor's cloud provider is a live risk whenever data traverses that provider, regardless of how quickly the vendor deletes it afterward.
2. Bots that auto-join calls
Notetakers that dial into Zoom or Teams as a visible participant create two problems. First, opposing counsel or a client can see "Otter Notetaker" or "Fireflies" in the participant list, which invites uncomfortable questions and can undermine client trust. Second, in one-party and two-party consent states, an unannounced bot recording a call may itself create a wiretap issue — the exact allegation at the heart of Brewer v. Otter.ai.
3. Consumer-grade DPAs (or none at all)
General-purpose SaaS notetakers often reserve real DPAs and BAAs for enterprise tiers. Solo and small-firm lawyers are typically on the plan that does not get the paperwork. Under Model Rule 1.6 and Opinion 512, the lawyer — not the vendor — carries the confidentiality duty, so the absence of a DPA is the firm's problem, not the vendor's.
4. Ambiguous training language
"We do not use your data to train third-party models" is not the same as "we do not use your data to train models." Read carefully. In some product tiers, content may be used to improve the vendor's own systems, human reviewers may sample transcripts for quality, or content may be shared with a model provider under that provider's own terms.
5. Speaker-ID marketing you cannot verify
Many tools advertise speaker identification. In practice, accuracy varies dramatically with microphone setup, accents, cross-talk, and the number of participants. Treat any speaker-labeled transcript as a draft that a human must verify before it is relied on for a deposition summary, witness prep, or a billing narrative.
What to Look For Instead
On-device processing
The cleanest way to answer the confidentiality question is to make sure the audio never leaves the lawyer's machine. On Apple Silicon Macs, the Neural Engine is fast enough to run high-quality speech-to-text and summarization locally. If nothing is uploaded, there is no server to subpoena, no subprocessor list to audit, and no training-data ambiguity to negotiate around. It is privilege-safe by architecture rather than by policy.
Capture without a bot
For virtual meetings, look for tools that capture system audio locally instead of joining as a participant. Nothing appears in the Zoom or Teams participant list; the recording exists only on the attorney's device. This does not eliminate consent obligations — you still need to comply with your jurisdiction's recording law and your client's expectations — but it removes the wiretap-adjacent risk of a hidden third-party bot.
Paperwork on the standard plan
A vendor that will sign a DPA and an NDA for a solo practitioner is a vendor that has thought about the legal market. A vendor that requires an enterprise contract before it will discuss confidentiality terms has not.
Matter-centric organization
General notetakers organize by "meeting." Firms organize by matter. Look for tools that let you tag every recording to a client and matter, log consent at capture, and label output as Privileged & Confidential — the working conventions that make an AI notetaker fit into an actual case file rather than an inbox.
Comparison: Cloud Notetakers vs. On-Device
| Dimension | Cloud notetakers (Otter, Fireflies, Fathom, tl;dv) | On-device (e.g., Basil) |
|---|---|---|
| Audio location | Uploaded to vendor cloud + subprocessors | Stays on the attorney's Mac |
| Server to subpoena | Yes — vendor and cloud provider | No vendor server involved |
| Training-data risk | Depends on tier and policy language | Vendor never receives content |
| Bot in the meeting | Common (visible participant) | None — local system-audio capture |
| DPA / NDA on solo plans | Often enterprise-only | Signed on request |
| Offline use | No | Yes |
| Matter tagging / privilege labels | Rare | Built for legal workflow |
How the Ethics Opinions Line Up
The two most-cited authorities on lawyer use of AI are converging on the same practical instruction: understand the tool, understand where the data goes, and do not treat vendor marketing as a substitute for diligence.
- ABA Formal Opinion 512 requires lawyers to evaluate a generative AI tool's confidentiality posture, obtain client-informed consent where the risk warrants it, and remain competent in the technology they use.
- NYC Bar Formal Opinion 2025-6 reinforces the confidentiality analysis and specifically flags the risk of feeding client information into third-party AI services without adequate protections.
- US v. Heppner illustrates the downstream litigation risk: a court can, and did, treat AI-platform interactions as non-privileged third-party disclosures.
- West Technology Group v. Sundstrom (D. Conn. 2024) is a useful reminder that inadvertent disclosure analyses turn heavily on the reasonableness of the precautions the disclosing party took.
Reading these together, the safer procurement posture is not "which vendor has the best privacy page" but "which architecture requires the fewest promises to trust."
Practical Selection Checklist
When you sit down to evaluate an Otter.ai alternative, work through this in order:
- Confirm where audio, transcript, and summary are processed and stored.
- Ask for the current subprocessor list in writing.
- Ask whether a DPA and NDA are available on your plan — not a higher one.
- Read the training-data clause and the human-review clause of the ToS carefully.
- Test the product on a mock matter, not a real client call.
- Document your diligence. Opinion 512 essentially requires it, and a short internal memo is cheap insurance.
- Update your engagement letter to describe how AI is used in the representation, and obtain consent where appropriate.
For more on the engagement-letter piece, see our companion piece on Basil for Law and the practical implications of ABA Opinion 512 for notetakers.
Where the Common Alternatives Land
A quick, factual read on the alternatives most firms consider when moving off Otter.ai:
- Fireflies.ai — cloud transcription, joins calls as a bot, offers enterprise-tier DPAs. Same architectural profile as Otter for confidentiality purposes.
- Fathom — cloud transcription, integrates tightly with Zoom, marketing emphasizes not training on customer data. Still a cloud service with subprocessors.
- tl;dv — cloud transcription, bot-based capture, marketed to sales and product teams rather than regulated professions.
- Microsoft Teams Premium / Copilot — recordings and transcripts live in the Microsoft cloud under the firm's tenant. Better contractual posture for many firms, but still a cloud pipeline with model providers in the loop.
- Basil — on-device on Apple Silicon Macs. No server, no subprocessors, no bot in the meeting. Designed around the confidentiality duty rather than retrofitted to it.
How Basil approaches this
Basil was built by a practicing lawyer for the specific problem this article describes. Audio capture, transcription, and summarization all run on the Apple Neural Engine — locally, on the attorney's Mac. Nothing is uploaded. There is no Basil server that receives client content, and there are no AI subprocessors in the path because the models run on-device. For virtual meetings, Basil captures system audio in "Computer mode" without joining the call, so no bot appears in the participant list.
The general Basil app is available today with a free tier (60 minutes per month). The Basil for Law edition — adding privilege attestation, a consent log, matter organization, and Privileged & Confidential labeling — launches in August 2026 at $19.99/month or $199.99/year for solos, with a 3-day monthly trial and a 7-day annual trial. Basil will sign a DPA and an NDA on request.
The point is not that Basil is the only defensible choice. It is that the question a firm should be asking a notetaker vendor is architectural, not aesthetic — and the answer should be verifiable without taking anyone's word for it.
This article is for information only and is not legal advice.
Frequently asked questions
Is Otter.ai safe for attorney-client conversations?
Otter.ai is a cloud service that uploads audio to its servers and involves subprocessors. That architecture creates confidentiality and third-party disclosure considerations that ABA Formal Opinion 512 and NYC Bar Opinion 2025-6 ask lawyers to evaluate carefully. Whether it is appropriate for a given matter depends on the firm's diligence, the client's informed consent, and the terms actually in place with the vendor.
What is the main difference between a cloud notetaker and an on-device notetaker?
A cloud notetaker sends audio to a vendor's servers for transcription and summarization. An on-device notetaker performs those tasks locally on the attorney's computer. The practical difference is that on-device tools do not create a vendor-side copy of the conversation, which reduces subpoena exposure and eliminates subprocessor risk.
Do I need client consent to use an AI notetaker?
ABA Formal Opinion 512 indicates that informed client consent is often appropriate when generative AI tools materially affect confidentiality. Best practice is to disclose AI use in the engagement letter, describe the tool's architecture, and obtain written consent. Jurisdictional recording laws also apply and are independent of the AI question.
Will an AI notetaker waive privilege?
No tool can guarantee that privilege will or will not be preserved — that determination is made by a court on the facts. However, US v. Heppner (S.D.N.Y. 2026) illustrates that disclosing content to a third-party AI service can be treated as a non-privileged third-party disclosure. Choosing an architecture that avoids third-party disclosure reduces that risk.
Does Basil join Zoom or Teams calls as a bot?
No. Basil captures system audio locally on the attorney's Mac in Computer mode. Nothing appears in the participant list of Zoom, Teams, or Meet, and no third-party bot is inserted into the meeting.
Does Basil sign DPAs and NDAs for solo attorneys?
Yes. Basil signs DPAs and NDAs on request, including on solo plans. Because processing is on-device and Basil never receives user content, the practical scope of those agreements is narrower than with cloud vendors.
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 → · Legal-tool reviews →
This article is for information only and is not legal advice.