Your AI note-taker is a POPIA problem

Zeldeen Müller, CEO, inSite Connect, explains why boards need to know exactly where meeting recordings, transcripts and sensitive information are being stored – and who ultimately controls that data.

2
Zeldeen Müller, CEO, AgendaWorx
Zeldeen Müller

Where does the meeting recording go to?

Somewhere in the last year, a new attendee started joining trustee meetings. Nobody voted it onto the board. It sits quietly, listens to everything and emails a neat summary afterwards. It is an AI note-taker, and most funds have no idea where the recording of their meeting goes.

The convenience is real. A tool like read.ai, Otter or Fireflies will transcribe a two-hour meeting, pull out the action items and save an evening of writing minutes. But a trustee meeting is not an ordinary call. It carries member data, legal opinions, actuarial figures and board resolutions, all protected under POPIA, and the governing body is accountable for keeping it safe.

Where the recording goes

This is the question to ask before the next bot joins. When you use an open AI service, every prompt and response travels to read.ai, Fireflies or Otter’s servers. On some of these, especially consumer and premium accounts, that content can be used to train the models by default, unless someone switches the control off. Even if training is switched off, the recording has been processed and stored outside the fund’s direct control.

There is no data processing agreement between the fund and the tool. There is no contractual right to audit what happens to the content afterwards. You are relying on terms the fund never negotiated and has no right to audit, with the very data you are duty-bound to protect. If a member asks where the recording of a discussion about their benefit was stored, most boards could not answer.

The closed alternative

The good news is that you do not have to give up transcription to stay safe. You must keep the recording inside an environment the fund controls. A closed system is one where the recording, transcript and summary stay within the fund’s perimeter.

Microsoft Teams, running inside your own Microsoft 365 tenant, can record and transcribe a meeting under terms your organisation has agreed. Better still, your secure board portal may already offer a closed AI transcription feature, where the meeting transcription is processed on a service provider’s server that operates under the fund’s control. In each case, the recording never leaves the fund’s control.

The difference is not the feature. Both the open tool and the closed one will hand you a tidy summary. The difference is everything sitting behind it. With a closed AI system, you get contracts, audit logs and regional controls to prove where the data lives, who touched it and assurances that it was never fed into someone else’s training run.

What a board should do

Start with a simple rule. No AI attends a trustee meeting unless the board knows exactly where the recording goes and can prove it. That means:

  • Naming the approved tool in your AI policy.
  • Switching off any personal note-takers that individual trustees or service providers might bring along.
  • Choosing transcription tools that are either in a closed AI system or that come with a data processing agreement and an audit trail.

If your board portal offers closed transcription, use it. If it does not, Teams inside your own tenant is a reasonable second choice. What you should not do is let an unvetted bot from a browser tab quietly record your most sensitive discussions.

The technology is genuinely useful. The real risk is discovering only afterwards that the most sensitive discussions of the fund were recorded and stored somewhere the board cannot control.

Stay safe out there,

Zeldeen Müller

Zeldeen Müller, CFP®, is a trailblazing entrepreneur with 26 years in the retirement fund industry. As CEO of inSite Connect and creator of AgendaWorx, she launched a secure, AI-powered board portal in 2011, revolutionising secure board management. A cybersecurity expert, she ensures robust data protection. Her team’s 23 awards showcase her leadership. Müller advocates using AI cautiously, urging financial advisors to encourage clients to have AI suggestions checked by them for accuracy.


 

Sponsored