Fathom AI Notetaker Review for Lawyers
Fathom is one of the most widely adopted AI meeting notetakers on the market, with a free tier that is unusually generous and a paid tier that has become a default in sales and customer-success stacks. Lawyers are increasingly asked whether it is safe to point at a client call, a deposition prep session, or an internal case strategy meeting. This review takes that question seriously.
Disclosure: Basil, which publishes this review library, competes with Fathom in the meeting-notetaker category. Basil is a fully on-device notetaker built by a practicing lawyer; see Basil for Law for how our architecture differs. We have tried to keep this review fair and to praise what Fathom genuinely does well.
What Fathom Actually Does
Fathom joins your Zoom, Google Meet, or Microsoft Teams call as a bot participant, records the meeting, transcribes it, and produces an AI summary with action items, timestamps, and topic segmentation. It integrates with common CRMs (Salesforce, HubSpot, Close) and pushes structured notes into Slack, email, and Notion. Users can generate follow-up emails, extract playbook moments, and search across meeting history. See the vendor's own overview at fathom.video and the features page.
The product is well made. Summaries are consistently coherent, the topic detection is better than most competitors, and the free tier really is free for unlimited recordings, which is unusual. For non-privileged internal meetings — team standups, vendor calls, marketing brainstorms — many lawyers will find it perfectly serviceable.
The concern is not usability. It is what happens to the audio and transcript after the call ends.
The Confidentiality Lens
This is the centerpiece of any legal-tech review, so it gets a careful walk-through.
Where processing happens. Fathom is a cloud service. Meeting audio is uploaded to Fathom's infrastructure, transcribed there (Fathom has historically used a mix of its own and third-party models), and stored on Fathom's servers. This is not a local-processing tool. See the Fathom security page and the privacy policy for the vendor's own description.
Sub-processors and model providers. Fathom's privacy documentation discloses use of third-party sub-processors, including large-model providers, to generate summaries and other AI outputs. Fathom states it has agreements requiring those providers not to train on customer data. That is a contractual protection, not an architectural one — meaning it depends on the sub-processor honoring the contract, and on the chain of custody staying intact.
Training use. Fathom's public position is that it does not use customer meeting content to train its own or third-party models. Confirm the current language on the privacy policy before onboarding; policies change.
Retention. Recordings and transcripts are retained in Fathom until the user or workspace admin deletes them. There is no default short-window purge for legal users. If you record a client call in January and forget about it, it is still on Fathom's servers in December.
DPA and BAA. Fathom offers a Data Processing Agreement for business and enterprise customers. It does not, as of this writing, publicly market a HIPAA BAA for meeting content, which matters if you handle PHI (personal-injury, medical-malpractice, ERISA disability, elder law). If HIPAA is in your practice, confirm current BAA availability directly with Fathom sales before use.
What a subpoena to Fathom could reach. This is the underappreciated risk. Because audio and transcripts sit on Fathom's servers, a third party in litigation who learns your firm uses Fathom can, in theory, serve a subpoena on Fathom for meeting content. Whether that content is privileged is a separate question the court will decide — but the vendor is a custodian of the record. Compare the reasoning in US v. Heppner (S.D.N.Y. Feb 2026), where Judge Rakoff held that a litigant's chats with a public AI platform were not privileged; and Brewer v. Otter.ai (2025), which put a spotlight on meeting-bot recording practices. The architectural reality is simple: data you send to a vendor is data a subpoena can chase at the vendor.
None of this makes Fathom disqualifying. It makes Fathom a tool whose use case for legal work needs to be scoped intentionally.
Ethics-Opinion Fit (ABA 512)
ABA Formal Opinion 512 (July 2024) is the current touchstone for generative-AI use by lawyers. It does not ban cloud AI tools. It requires competence about how the tool works, informed client consent where confidential information is involved, reasonable measures to prevent inadvertent disclosure, and careful supervision of AI output. NYC Bar Formal Opinion 2025-6 (Dec 2025) reinforces the same themes for New York practitioners, and Model Rule 1.6 supplies the underlying confidentiality duty.
Applying that framework to Fathom:
- Competence. You need to understand that Fathom is cloud-based, uses sub-processors, and retains content by default. That is fine — you just need to actually know it.
- Client consent. If a Fathom bot will be present on a call involving client information, opinion 512 pushes toward informing the client. That is not merely notifying the other participants that a recording is happening (which Fathom does with an on-screen indicator); it is a separate conversation with your client about the tool.
- Reasonable measures. Turning on SSO, enabling admin controls, disabling auto-record for legal seats, and setting a retention/deletion policy are the baseline moves.
- Supervision. AI summaries are drafts. Fathom summaries are good, but they hallucinate action items, misattribute statements, and occasionally invent nuance that was not said. Review before circulating.
Fathom is compatible with opinion 512 for many legal use cases. It is not the lowest-risk architecture available for the most sensitive ones.
Where Fathom Fits, and Where It Does Not
Reasonable fits: internal firm meetings that do not discuss client matters; business-development calls; vendor demos; recruiting interviews; CLE prep; podcast recordings; non-privileged intake conversations where the prospective client has been informed and consented.
Harder fits: privileged strategy sessions with co-counsel; witness prep; settlement calls; conversations with clients about ongoing litigation; anything covered by a protective order; matters involving PHI without a BAA in place; matters where opposing counsel is aggressive about third-party discovery.
For those harder categories, an on-device notetaker keeps the recording, transcript, and summary on the lawyer's own machine, which materially shrinks the vendor-subpoena surface. That is the design philosophy behind Basil for Law. Whichever tool you choose, the architectural question — cloud or local — should be a conscious decision, not a default.
Pricing and Who It's For
Fathom's free tier is genuinely useful and has been a major driver of its adoption. Paid tiers add team features, CRM sync, advanced AI actions, and admin controls. Because pricing is updated periodically, refer to the current Fathom pricing page rather than a stale number here.
The tool is aimed squarely at sales, customer success, and product teams. Legal is not its primary market, which shows up in things like the lack of matter-scoped retention, the absence of prominent BAA marketing, and the general assumption that meetings should be indexed and searchable across a workspace. None of that is wrong for its target audience. It is just worth naming when you are evaluating fit.
Verdict
Fathom is a very good general-purpose notetaker. For legal work, whether it belongs in your stack depends on the sensitivity of the meetings you want to capture and how disciplined you are about scoping which calls the bot joins. It is a real tool that does real work, and it should be evaluated on that basis rather than dismissed. It should also not be treated as a default for privileged conversations.
| Pros | Cons |
|---|---|
| Excellent summary quality and topic detection | Cloud architecture: audio and transcripts sit with the vendor |
| Generous free tier; low friction to try | Sub-processors involved in AI summarization |
| Strong integrations (Zoom, Meet, Teams, CRMs) | No prominent HIPAA BAA marketing for meeting content |
| DPA available for business customers | Default retention favors indexing over minimization |
| Clear recording notification to participants | Vendor-held records are reachable by third-party subpoena |
This review is for information only and is not legal advice.
Frequently asked questions
Is Fathom safe to use for privileged client meetings?
Fathom is a cloud service that stores audio and transcripts on its servers and uses third-party sub-processors for AI summarization. That is not automatically disqualifying, but for privileged conversations you should consider client consent under ABA Formal Opinion 512, a Data Processing Agreement with Fathom, retention controls, and whether an on-device alternative would materially reduce the vendor-held data surface. It is a risk-reduction decision, not a yes-or-no answer.
Does Fathom train AI models on my meetings?
Fathom's public position is that it does not use customer meeting content to train its own or third-party models, and it states that its AI sub-processors are contractually barred from doing so. Because policies change, confirm the current language on Fathom's privacy policy before onboarding and again periodically.
Can a subpoena to Fathom reach my client meetings?
In principle, yes. Any cloud vendor that holds recordings and transcripts is a potential custodian and can be served with a subpoena. Whether the underlying content is privileged is a separate question the court decides, but the practical exposure exists whenever data leaves your device. This is the core architectural tradeoff between cloud and local-first notetakers.
Does Fathom offer a HIPAA BAA?
Fathom does not prominently market a HIPAA Business Associate Agreement for meeting content as of this writing. If your practice regularly involves PHI, confirm current BAA availability with Fathom sales directly before using it on any call that could involve protected health information.
How does Fathom compare to Basil for legal work?
Fathom is cloud-based and optimized for sales and CS teams; Basil is an on-device notetaker built by a practicing lawyer, where audio, transcript, and summary stay on the user's machine. The two tools solve overlapping problems with different tradeoffs on convenience, sharing, and where client data physically lives. See Basil for Law for the architectural details.
What should I do at minimum before using Fathom on any legal-related call?
At minimum: sign the Fathom DPA, enable SSO and admin controls, turn off auto-record for legal seats, set a written retention and deletion policy, decide which meeting types the bot may join, obtain informed client consent where confidential information will be discussed, and always review AI-generated summaries before circulating them.
Meeting notes with no server to subpoena
Basil transcribes and summarizes entirely on-device — privilege-safe by architecture. See Basil for Law →
This review is for information only and is not legal advice.