AI Meeting Notes Privacy: What Google Meet’s Change Means

AI meeting notes privacy has moved from a niche legal worry to a practical question every organisation must answer. On 29 September 2026, new Google Meet settings that pre-configure its “Take notes for me” feature begin taking effect, switching automatic AI note-taking on by default for some business plans — just weeks after a US federal court let most of a major privacy class action against Otter.ai proceed.

For Indian companies, training institutes and agencies that record and summarise meetings every day, these two developments are a timely prompt to ask a simple question: who controls the notes an AI writes about your conversations?

What changed in Google Meet on 29 September 2026

Google first announced the new controls in a July 2026 Workspace Updates post. Until now, administrators could only switch automatic note-taking fully on or fully off. The update adds a third choice: turn on automatic notes only for meetings with three or more people. End users get a matching personal option.

The defaults are what make this newsworthy. According to Google’s dedicated 29 September rollout notice, the new setting is ON by default for Business Standard and Business Plus, while it is OFF by default for Enterprise Standard, Enterprise Plus, Frontline Plus and the Google AI Pro for Education add-on. Google’s weekly recap of 25 September confirmed the settings “begin to take effect” on 29 September, and advised admins to review their configuration beforehand. Organisations that tested the feature through the Gemini Alpha programme were warned it may already be switched on.

In other words, many small and mid-sized businesses on business tiers could find an AI quietly generating notes in group calls unless someone in IT deliberately checks the setting. That is not a scandal — it is a product decision — but it illustrates how easily AI transcription becomes the default rather than a conscious choice.

Why AI meeting notes privacy is now a legal issue

The Google change lands in a year when AI notetakers have been tested in court. In the consolidated case In re Otter.ai Privacy Litigation in the Northern District of California, the court on 13 August 2026 allowed significant portions of the proposed class action to proceed, including claims under the federal Electronic Communications Privacy Act, California’s Invasion of Privacy Act and Illinois’ Biometric Information Privacy Act, as summarised by Hintze Law on 3 September 2026. Otter denies wrongdoing and no court has ruled on the merits.

UC Today’s analysis of the case explains the core allegation: that an AI assistant joined and transcribed Zoom, Teams and Meet calls without the consent of every participant. Otter’s position is that it did not intercept communications unlawfully.

The litigation is spreading. A Computerworld report from 9 September 2026 lists suits against Fireflies over voiceprints under Illinois biometric law, a Washington case over Microsoft Teams live transcription, and a newer claim against Granola alleging it was designed to be used without participants knowing and trained AI on conversation data. Brian McGinnis, a partner at Barnes & Thornburg, told the publication that companies deploying these tools across jurisdictions “have to play to the most stringent state’s law.”

These are US cases, but the underlying questions are universal:

• Consent — did everyone in the meeting know an AI was listening and writing things down?
• Purpose — is the transcript used only for minutes, or also to train someone else’s model?
• Storage — where do recordings, transcripts and summaries physically live, and for how long?
• Access — who inside and outside the organisation can read them?
• Biometrics — are voiceprints or face data being derived and retained?

The India angle: DPDP rules and meeting data

India has its own framework to think about. The Digital Personal Data Protection Rules were notified on 14 November 2025, with compliance phased in over 18 months so that core obligations apply by May 2027, as India Briefing’s guide to the DPDP Rules explains. Under that regime, organisations acting as data fiduciaries must give clear notices, obtain purpose-specific consent that can be withdrawn, maintain reasonable security safeguards, and report personal data breaches to the Data Protection Board within 72 hours. Penalties for failing to protect data can reach ₹250 crore.

A meeting recording is personal data. A transcript that names a candidate, a patient, a student or a client is personal data. An AI-generated summary that captures who said what is personal data. If that material sits on a third-party platform, is processed by an AI vendor you did not choose, or is shared by default with a wider group than intended, your compliance team will need answers well before May 2027.

This is why meeting transcription compliance is becoming an architecture decision, not just a policy document. Policies tell people what to do; architecture decides what is even possible.

Practical steps to protect AI meeting notes

Whatever platform you use, a few habits dramatically reduce risk:

1. Make AI note-taking a deliberate choice. Check your admin console today. If you are on a Google Workspace business tier, decide consciously whether automatic notes should be on, off or limited to larger meetings.

2. Announce it, every time. Tell participants at the start that the meeting is being recorded and summarised, and give them a genuine chance to object. The US lawsuits largely turn on people who say they were never told.

3. Limit who receives the notes. Summaries emailed to every invitee, including external guests, can spread sensitive content far beyond the room. Default to the smallest sensible audience.

4. Keep sensitive meetings out of scope. Many law firms advise barring AI notetakers from interviews, disciplinary meetings, medical consultations and internal investigations unless there is explicit consent and a clear need.

5. Know where the data lives. Ask your vendor exactly which cloud, which region and which AI provider process your audio and text — and whether any of it is used for model training.

6. Set a retention period. Transcripts kept forever are liabilities waiting to happen. Decide how long recordings and notes are genuinely needed, then delete them automatically.

How InstaDataTalk keeps meeting AI under your control

InstaDataTalk was built around the idea that organisations should own their meetings end to end. It is a complete white-label video meeting platform that you deploy on your own infrastructure, so recordings, files and transcripts are stored in your own cloud storage rather than on a shared public platform.

Several design choices speak directly to the risks above:

• Self-hosted by design. Meetings run on your servers under your own domain over HTTPS. Nothing is mined or resold, which supports data sovereignty for regulated conversations.
• Your storage, your rules. Recordings go to S3-compatible storage you control (for example Backblaze B2), with an automatic media-retention lifecycle so old files do not pile up indefinitely.
• Your AI keys. Meeting summaries and minutes come from a dual-AI engine — Claude via OPUSMAX as primary with automatic OpenAI fallback and a pre-flight health check — and you can use your own AI provider keys, so you know exactly which provider processes your content.
• Live on-device transcription. Captions and transcription are generated on the device, creating a searchable record without extra plug-ins.
• Privacy-safe reports. Attendees receive the summary and minutes as a secure report link, while engagement and attentiveness analytics stay private to the organiser. Proctoring and attentiveness features are consent-based.
• Admission control. Lobby, waiting room, mute-on-entry and permission settings help hosts decide who joins before any recording begins.
• No per-seat lock-in. There is no per-seat or per-minute metering, so moving sensitive meetings in-house does not multiply your licence bill.

Self-hosting does not remove your legal duties — you still need notices, consent and sensible policies. What it changes is control: you decide the defaults, the storage location, the AI provider and the retention period, instead of inheriting them from a vendor’s product roadmap. You can read more about this approach in our earlier pieces on the InstaDataHelp blog.

Frequently Asked Questions

What changed in Google Meet’s “Take notes for me” on 29 September 2026?

Google added an option to enable automatic AI note-taking only for meetings with three or more people, for both admins and end users. The setting is on by default for Business Standard and Business Plus, and off by default for Enterprise Standard, Enterprise Plus, Frontline Plus and Google AI Pro for Education customers.

Why is AI meeting notes privacy a legal risk?

In the US, AI notetaker vendors face lawsuits alleging they recorded or transcribed participants without consent and used conversations to train AI. In August 2026 a federal court allowed major parts of the Otter.ai class action to proceed. The claims remain unproven, but they show regulators and courts are watching.

Does India’s DPDP Act apply to meeting recordings and transcripts?

Recordings, transcripts and summaries that identify people are personal data. Under the DPDP Rules, whose core obligations apply by May 2027, organisations need clear notices, purpose-specific consent, security safeguards and 72-hour breach reporting to the Data Protection Board.

How does a self-hosted meeting platform help with privacy?

A self-hosted platform such as InstaDataTalk keeps recordings, transcripts and files in storage you control, lets you choose your own AI provider keys, and applies an automatic retention lifecycle, giving you direct control over where meeting data lives and how long it is kept.

Sources

• Google Workspace Updates: new Meet “Take notes for me” settings for admins and end users (July 2026)
• Google Workspace Updates: “Take notes for me” settings take effect 29 September 2026
• Hintze Law: ten ways to avoid privacy risks revealed in AI notetaker suits
• Computerworld: AI notetakers at work could leave companies at risk for lawsuits
• UC Today: Otter.ai on trial, and the AI notetaker industry with it
• India Briefing: DPDP Rules 2025 notified — compliance guide

Want AI meeting summaries without handing your conversations to someone else’s cloud? Explore InstaDataTalk or write to info@instadatahelp.com to set up self-hosted, white-label meetings on your own servers.