Is Otter AI Safe for Work? AI Note Taker Security for IT Teams

Short answer: Otter.ai can be used safely at work, but not by default and not by blocking a website. The real risk from AI note takers like Otter lives in the calendar and meeting permissions employees grant on their own, not in network traffic, so the first control is finding those OAuth grants in your Microsoft 365 or Google tenant. dope.security does that with AI-powered SSPM, then adds web policy and on-device DLP for the parts that do run on the laptop.

That framing matters because most "is Otter AI safe" advice stops at "read the privacy policy." IT teams need something more practical: where the data goes, who agreed to it, and which control can actually see it. We covered the general version of this problem in how AI-powered SSPM finds and scores every OAuth-connected app in your tenant. This post applies it to the most common shadow AI app in any meeting-heavy company.

We will cover what Otter does with meeting data, what the 2025 class action alleges and where it stands, why meeting bots slip past traditional security tools, and a control stack that works. For a broader map of how this pattern shows up, see our 12 real workplace shadow AI scenarios and fixes.

What Otter.ai does with your meetings

Otter is an AI meeting assistant. It records meetings on Zoom, Microsoft Teams and Google Meet, transcribes them, writes summaries and action items, and stores the results in the user's Otter account. Its assistant, historically called OtterPilot and now Otter Notetaker, can join meetings on its own when a user connects their calendar.

That auto-join behavior is the key detail for security teams. Once one employee links a work calendar to a personal or team Otter account, the bot can join any meeting on that calendar. It does not need the meeting host's approval, and it does not need anyone else in the meeting to have an Otter account. Several university IT departments publish instructions telling staff how to turn auto-join off in Otter's account settings, which tells you how often it surprises people.

Otter is also far from the only option. Zoom, Microsoft Teams and Google Meet all offer built-in AI summaries now, and a long tail of startups sells note takers with their own bots. That is good news for productivity and bad news for visibility: every one of them asks for the same calendar and meeting access, and most can be connected by an individual employee in under a minute.

The data that leaves is not a file. It is the full audio of a conversation, plus a searchable transcript, plus a summary. In a sales call that might be pricing. In a board prep call it might be M&A plans. In a healthcare company it might be patient details.

Is Otter AI safe? What the lawsuit alleges

In August 2025, plaintiffs filed a class action against Otter.ai in the Northern District of California, now consolidated as In re Otter.AI Privacy Litigation (No. 5:25-cv-06911). The complaint alleges that OtterPilot joined meetings through calendar integrations and recorded, transcribed and retained conversations without the consent of all participants, and that Otter used recordings to train its models. These are allegations, not findings.

On August 13, 2026, the court granted Otter's motion to dismiss in part and denied it in part. According to the court record as summarized by AI Lawsuit Tracker, claims under the federal wiretap statute (ECPA), the California Invasion of Privacy Act, Illinois BIPA, the UCL and unjust enrichment proceed, while the CFAA and several other claims were dismissed, many with leave to amend. The case is active.

For an IT or security leader, the legal outcome matters less than the question it raises: if a bot joins your meeting through one employee's calendar, who consented on behalf of everyone else? Your company may be the one that has to answer that question to a customer, a regulator or an auditor, regardless of how the case ends.

Why meeting bots slip past your security stack

Most security tools were built to watch people and devices. A meeting bot is neither. Here is where each common control stands.

• DNS filtering: it can block otter.ai in the browser, but the bot joins from Otter's cloud, so blocking the domain on laptops does not stop it from entering a meeting.
• Cloud proxy SWG: same limitation. It sees users browsing to Otter, not Otter's servers connecting to Zoom or Teams.
• Endpoint DLP: useful when someone uploads a recording or pastes a transcript, but the live meeting audio never passes through the user's device on its way to Otter.
• Meeting platform settings: Zoom, Teams and Meet have controls for external participants and apps, but they vary by platform and by who hosts the meeting.
• The OAuth grant: this is the one place the relationship is visible. To auto-join, Otter needs calendar access granted through Microsoft 365 or Google Workspace. That consent is recorded in your tenant.

That last line is the thesis of this post. The durable control point for AI note takers is the identity layer, specifically the third-party app permissions users approve, because that is where the bot gets its invitation list.

How to find Otter and other AI note takers in your tenant

Start with an inventory. In both Microsoft Entra ID and Google Workspace, admins can review third-party apps that users have authorized and the scopes they hold. The manual version works for a one-time check. The problem is that new AI note takers appear every month, and a list of app IDs does not tell you which grant is risky.

dope.security's AI-powered SSPM, part of CASB Neural, automates this. It inventories every app authorized on your Microsoft 365 or Google tenant with its scopes, risk rating, consent type, permissions used, sign-ins and status. It rolls up critical and high-risk apps, unused permissions, unverified publishers and apps missing an owner, and it writes a plain-language summary and recommended actions for each app. A calendar-read grant held by a meeting assistant that nobody in IT approved is exactly the kind of finding it surfaces.

We go deeper on the Microsoft side in how to find OAuth-connected shadow AI apps in Microsoft 365, including what to look for in consent type and publisher verification.

A five-step control stack for AI note takers

No single setting fixes this. The goal is to decide which note takers you sanction, make the approved one easy, and make the unapproved ones visible and costly to use.

• 1. Inventory the grants. Use SSPM or your IdP's app consent report to find every meeting assistant with calendar or mail scopes. Note who granted it and when.
• 2. Pick a sanctioned tool. Choose one note taker (your meeting platform's built-in AI, or an enterprise plan of a third-party tool) and confirm retention, training and consent settings in the contract.
• 3. Tighten user consent. Require admin approval for apps requesting calendar or mail access, so new bots cannot be connected silently.
• 4. Set web policy. Use your SWG to allow the sanctioned note taker and block or warn on others, so employees get a nudge at the moment they try to sign up.
• 5. Watch the files. Transcripts and recordings get exported and uploaded elsewhere. On-device DLP that reads uploads by meaning catches the moment a transcript full of customer data heads to a personal drive or another AI tool.

Steps 4 and 5 only work if policy changes reach every laptop fast. When Outreach Health moved web policy onto the device with dope.security, policy changes went from days to minutes and web-access tickets fell 70 percent in 90 days. That speed is what lets you respond the week a new note taker goes viral inside your company.

Write the policy down too. Our AI usage policy template with enforcement steps includes language for recording and transcription tools.

SSPM plus on-device controls vs a block-the-domain approach

Teams usually reach for one of two approaches. Here is how they compare in practice.

Blocking the Otter domain

Blocking otter.ai at DNS or in a cloud proxy stops employees from using the Otter web app on managed laptops. It does not remove calendar grants that already exist, it does not stop the bot from joining meetings from Otter's cloud, and it does not catch the next note taker with a different domain. It feels decisive and leaves the actual data path open.

The dope.security approach

dope.security starts where the bot gets its access. CASB Neural's SSPM finds the OAuth grant, scores it and recommends the fix. On the laptop, the Fly-Direct SWG enforces allow, warn or block by category or custom category, with policy that lands on every endpoint in under a minute. AI Analytics shows which AI apps employees actually use, who uses them and how much data moves, including personal versus enterprise-licensed accounts for tools like ChatGPT and Claude. Dopamine DLP classifies uploads by meaning on the device before they leave.

• Existing grants: domain blocking leaves them in place; dope.security SSPM surfaces them with a risk rating and recommended action.
• New note takers: domain blocking needs a new rule per tool; dope.security AI Analytics discovers AI apps as they appear, sanctioned or not.
• Transcript leakage: domain blocking cannot see it; dope.security Dopamine DLP blocks sensitive uploads to the web and AI tools from the endpoint.
• User experience: domain blocking is a dead end; dope.security can warn instead of block, so users get a nudge rather than a dead end.

What to ask any AI note taker vendor

If you decide to sanction a note taker, put these questions in the security review. They map directly to the allegations in the Otter case and to common audit findings.

• Does the bot announce itself to every participant, and can admins force that behavior on?
• Can admins disable auto-join organization-wide, or only users individually?
• Are recordings or transcripts used to train models, and is that opt-in or opt-out at the org level?
• What are the default retention periods, and can admins enforce shorter ones?
• Does the enterprise plan support SSO, SCIM and domain capture so personal accounts can be consolidated?
• Where is data stored, and what certifications cover it?

Get the answers in the contract. A feature on a roadmap slide does not help you in a deposition.

The bottom line

Otter is not unsafe by nature, but an unmanaged AI note taker is an uncontrolled recording device with a calendar invite to your most sensitive meetings. The fix starts at the OAuth grant, not the firewall, and finishes on the device where transcripts get shared. That is the order dope.security works in: CASB Neural with AI-powered SSPM to find the grants, and the Fly-Direct SWG with Dopamine DLP to govern what happens next. Book a demo and we will show you every AI app connected to your tenant in one screen.

Frequently Asked Questions

Is Otter AI safe to use for work meetings?

It can be, if IT sanctions it on an enterprise plan with known retention, training and consent settings. The risk comes from unmanaged accounts whose bot auto-joins meetings through an employee's calendar. dope.security helps by finding those calendar grants with AI-powered SSPM and governing web access and uploads on the device.

Can Otter join meetings without the host's permission?

When a user connects a calendar and enables auto-join, Otter's assistant can join meetings on that calendar as a participant. The 2025 class action alleges this happened without all participants' consent, and Otter disputes those allegations. Users can turn auto-join off in Otter's settings, and admins can restrict which apps may access calendars.

What is the Otter.ai lawsuit about?

In re Otter.AI Privacy Litigation, filed in August 2025 in the Northern District of California, alleges Otter recorded and retained meeting conversations without all participants' consent and used them for model training. On August 13, 2026, the court let wiretap, CIPA, BIPA and related claims proceed while dismissing others. The case is ongoing.

Will blocking otter.ai on our network stop the bot?

No. The bot joins meetings from Otter's cloud, so blocking the domain on laptops stops the web app but not a bot already connected to a calendar. Remove or restrict the OAuth grant in Microsoft 365 or Google Workspace first, then use web policy for new sign-ups.

How do I find which employees use AI note takers?

Review third-party app consents in Microsoft Entra ID or the Google Workspace admin console for apps with calendar or mail scopes. dope.security automates this: SSPM lists every authorized app with scopes, consent type and risk, and AI Analytics shows which AI apps are used on endpoints and by whom.

Can DLP stop meeting transcripts from leaking?

DLP cannot see live meeting audio flowing from Zoom or Teams to a vendor's cloud. It can stop what happens next, when someone exports a transcript and uploads it to a personal drive or another AI tool. dope.security's Dopamine DLP classifies those uploads by meaning on the device and can block them.