AI scribes are embedded in all matter of channels such as Zoom and Google Meet. There can be benefits, but for those concerned about legal issues, estate structure and similar matters, there is now a governance headache.

The following article touches on a topic that the editor of Family Wealth Report has discussed before: what happens when AI devices are part of boardroom meetings and other discussions? (In the case of the FWR editorial, it said there is a risk that client confidentiality and privilege is destroyed.) 

The author of the following article is Jay Rogers (more details below), who has written for this news service before. The editors are pleased to share these insights; the usual editorial disclaimers apply. To comment, email tom.burroughes@wealthbriefing.com and amanda.cheesley@clearviewpublishing.com

Most family offices didn’t decide to start keeping a verbatim record of every investment committee meeting. The decision was made for them, by a default setting that nobody changed.

AI scribes are now embedded in Zoom, Microsoft Teams, and Google Meet. A parallel class of autonomous bots, such as Otter.ai, Fireflies, and similar tools, syncs to a calendar, joins any scheduled call as a visible participant, and produces a transcript, a speaker-attributed summary, and an action-item list without anyone in the room doing anything beyond keeping the defaults. For a routine portfolio update, that’s a productivity gain. For an investment committee meeting where estate structure, liquidity, succession, or legal strategy is on the agenda, it’s a governance problem that the family didn’t know it had.

Handwritten notes were impermanent, idiosyncratic, and hard to produce in a form that held up in discovery. AI transcripts are durable, searchable, speaker-attributed, and held indefinitely on third-party servers. The tool that saves time also manufactures the most complete record of a family’s private deliberations that most families have ever kept, whether or not they meant to.

Three mechanisms that create exposure
The first is attorney-client privilege. Privilege is generally waived when a confidential communication is voluntarily disclosed to a third party. Whether an AI transcription vendor qualifies as a protected participant, the functional equivalent of a stenographer, or counts as a third party whose presence destroys the protection is an unsettled legal question, the answer may depend on jurisdiction and how the vendor relationship is structured. Family office meetings often blend investment decisions with estate, tax, and legal strategy, with counsel in the room for both. Running an AI scribe through that session isn’t a neutral productivity choice; it’s an unresolved privilege question. The time to resolve it is before the recording starts, not after a dispute arises.

The second is biometric liability. Speaker-identification features in some AI tools analyze voice characteristics to attribute speech to named individuals, a process that can generate a “voiceprint” regulated under the Illinois Biometric Information Privacy Act and similar statutes in other states. Class actions against Otter.ai and Fireflies.ai allege the tools built and retained voiceprints without providing notice or obtaining consent. Critically, exposure doesn’t stop at the software vendor. Courts have found that the organization deploying the tool can be liable alongside it. Statutory damages run $1,000 to $5,000 per violation and require no proof of actual harm. A single meeting with a participant in a covered state is enough to trigger the analysis.

The third is discovery. Estate disputes, divorce proceedings, and intra-family litigation are among the most common legal contests a family office faces. In any of those proceedings, a time-stamped, speaker-attributed transcript of the meeting where the family discussed liquidity, asset values, a forthcoming distribution, or a change in beneficiary treatment is exactly what opposing counsel requests first. AI records are permanent and specific in ways that informal notes never were. That specificity cuts against the family when the transcript captures a remark that was exploratory, not final.

What good governance of these tools looks like
A well-governed AI notetaker policy answers four questions before a meeting begins: who may activate the tool, who must consent, what happens when counsel is present, and where the transcript lives afterward.

Consent and activation. The policy should name who has authority to enable a scribe on a given call and establish all-party consent as the default, regardless of where participants are physically located. States including California, Illinois, Florida, and Massachusetts require unanimous consent to record; a bot that joins and records without explicit opt-in from every participant risks wiretap liability in those jurisdictions. The simplest approach is a standing disclosure sent to all expected attendees before any regularly scheduled meeting, with an explicit opt-in mechanism for external participants joining for the first time.

A hard stop for legal sessions. Any meeting segment where counsel is present for estate, tax, litigation, or deal strategy should run without the scribe, or move to a channel the vendor doesn’t touch. This isn’t a judgment call left to the meeting host; it should be written into the policy as a standing rule. The productivity cost of pausing or switching is negligible. The cost of a waived privilege argument in active litigation is not.

Vendor diligence as an annual item. The terms of service for every AI tool used in investment committee or planning meetings should be reviewed annually for three things: how long transcripts are retained and where they’re stored, whether captured audio is used to train the vendor’s models (at least one major platform retains this right unless users opt out of a buried account setting), and what the vendor’s breach-notification and data-deletion protocols are. Most family offices have never read these terms. Reviewing them takes an afternoon and produces a documented record of due diligence.

The transcript is not the minutes. Official meeting minutes should be produced by a human reviewer working from the transcript as a draft, edited, approved by the principals, and distributed as the governing record of what the committee decided. When a vendor’s auto-generated summary becomes the default record by inertia, the family has handed control of its own governance documentation to an algorithm. The approved minutes are what get produced in a dispute. The raw transcript, if it also exists, gets produced alongside them.

The governance gap is a policy decision
Family offices that have worked through this framework report a consistent outcome: separating the AI-assisted operational portion of a meeting from the counsel-present planning discussion, combined with a clear approval process for official minutes, produces a cleaner record with less exposure and no meaningful loss of the productivity the tool provides. The scribe runs for the portfolio update. It stops before the estate conversation starts. The minutes reflect what the committee decided, not everything that was said on the way to deciding it.

The AI notetaker isn’t going away and governing it well doesn’t require abandoning it. It requires deciding, in writing, before the next meeting, who’s in control of the record. Right now, for most family offices, that decision hasn’t been made. The tool has.

About the author
Jay Rogers is president of Alpha Strategies Investment Consulting, a retained expert witness in securities and fiduciary litigation, and a financial professional with more than 30 years of experience in private equity, private credit, hedge funds, and wealth management. He has a BS from Northeastern University and has completed postgraduate studies at UCLA, UPENN, and Harvard. He writes about issues in finance, constitutional law, national security, human nature, and public policy.