AI meeting notetakers taking over your organization, and how to take back control

That’s exactly what happened at one of our customers, a healthcare company. Four employees started using Read.AI, a tool that joins meetings, records and summarizes them. Before long it was showing up in almost every meeting in the organization, including meetings other managers had set up. The customer’s request was simple: let the tool work for those four users, and make it disappear for everyone else.

Read.AI is just one example. Otter, Fireflies, Fathom and dozens of other tools work the same way. This article explains how a tool like this gets in without approval, why it’s risky, and how to take back control. It’s one of the most common cases we run into in our Microsoft 365 security projects.

How it gets in without anyone approving it

There’s no breach and no bug here. Everything works exactly as designed:

Sign-up. An employee signs up with their Microsoft 365 account, or clicks a link in a meeting summary the tool sent. That creates an account in the tool and connects it to their work account.
Consent. The sign-in screen asks for permissions: calendar, profile and sometimes email. If your organization lets users approve apps themselves (user consent), the employee clicks “Accept” and that’s it. IT isn’t involved.
Auto-join. The tool reads the employee’s calendar and joins meetings with a Teams link. Depending on the auto-join settings chosen at sign-up, it can join every meeting on the calendar, including ones the employee was only invited to, and sometimes even if the employee never joins.
Spread. After the meeting, the tool sends a summary to every attendee with an invitation to sign up. Each attendee who signs up joins the circle, and the tool spreads from meeting to meeting.

In Teams the tool joins as an anonymous participant. Until recently that was enough to get into meetings at most organizations. Today Teams detects most of these bots and holds them in the lobby by default, but the organizer can still admit them, and detection doesn’t catch every bot.

Why it’s risky

A meeting notetaker isn’t just a nuisance. It records and transcribes the most sensitive conversations in your organization and stores them outside it:

Data leaves the organization: recordings, transcripts and summaries live in a third party’s cloud under its terms, not under your retention, encryption and DLP policies.
Meetings nobody meant to share: board meetings, HR conversations, legal discussions and customer calls can all be recorded just because one invitee connected the tool.
Privacy and consent: people are recorded without being asked, which raises privacy and regulatory questions, especially for organizations holding medical or personal data.
Permissions that linger: even after the employee stops using the tool, its calendar access stays valid until someone explicitly revokes it.
Shadow IT: nobody knows how many of these tools are connected, what data they can reach, or who approved them.

That doesn’t mean AI notetakers are off limits. It means the decision should be the organization’s, not each employee’s.

How to take back control

For this customer, the goal was to keep the tool for four users and remove it for everyone else. The same steps work if you want to block a tool like this entirely:

Step	Where	What it does
Map the app and who approved it	Entra ID > Enterprise applications	Shows which users consented and which permissions the app received
Limit it to a group	In the app: Assignment required = Yes, assigned only to the approved users’ group	Only group members can sign in to the tool, and other users can no longer consent to it themselves. The approved users’ existing consents stay
Revoke everyone else	Remove the consents of users not on the list, through Microsoft Graph or PowerShell (the portal only shows them)	The tool loses access to their calendars and is no longer sent to meetings through their accounts. It can still arrive through another attendee who uses it
Block the app in Teams	Teams admin center > Manage apps	Blocks the tool’s Teams app for people who don’t need it. It doesn’t stop the bot joining through the meeting link
Control bots in meetings	Teams meeting policy: Manage external bots and their access to meetings, and Anonymous users can join a meeting unverified turned off	Detected external bots wait in the lobby for approval. These settings apply only to meetings your organization hosts
Settings in the tool itself	For approved users: Auto-Join Preferences set to meetings they host only	The tool stops joining other people’s meetings

It’s also worth telling employees what happened, which tools are approved, and what to do when an unknown tool shows up in a meeting: remove it from the participant list.

How to prevent the next one

Dealing with one tool solves today’s problem. To stop it coming back with the next tool, change the default:

Restrict user consent: in Entra ID, under User consent settings, choose Allow user consent for apps from verified publishers, for selected permissions, and classify only basic permissions as low impact. Calendar and email access can no longer be granted with one click. The change applies to future consents only, so review existing ones separately.
A proper approval process: turn on the admin consent workflow. An employee who needs an app sends a request with a reason, and an admin approves or denies it. People aren’t blocked, but the decision goes through IT.
An approved tools list: decide which AI tools are allowed and publish the list. If you have Microsoft 365 Copilot licenses, its meeting recaps stay within the Microsoft 365 service boundary and inherit your security, compliance and retention policies.
Monitor OAuth apps: in Defender for Cloud Apps, App Governance alerts you to new apps with broad permissions and lets you review every app already connected.
Regular review: once a quarter, go through Enterprise applications and check which apps received consent, how many users are connected, and which are no longer used.

What an app can actually do once it has consent, and why admin consent doesn’t give users new powers, is covered in our guide to Microsoft Graph permissions.

The bottom line

AI meeting notetakers get in through a single user consent and spread through the calendar from meeting to meeting. The answer isn’t to chase each tool separately but to change the default: restrict user consent, turn on an approval process, control bots in Teams meetings, and monitor which apps are connected to your tenant.

Want to know which apps are already connected to your environment? Book a call with the OnCloud team and we’ll review your consent settings, connected apps and meeting policies together.

Book a call