What is the current litigation risk with Otter.ai?

Judge Eumi K. Lee consolidated multiple suits into In re Otter.AI Privacy Litigation. Plaintiffs allege the AI Notetaker recorded meetings without all-party consent, with consent obtained at most from the meeting host. The motion to dismiss was heard May 20, 2026; no ruling had issued as of publish. The immediate exposure sits with the default that auto-joins forwarded calendar invites; buyers in the twelve all-party-consent states (covering roughly 35% of the U.S. population) inherit vicarious risk when their employees use the tool.

What retention windows should we set?

As a Cardinal judgment call — not a regulatory floor — a defensible starting point is audio at 30 days or less, transcript at 90 days or less, and summary at 12 months or less, with any override requiring explicit business justification. The point of writing it down is not the specific number; it is that a retention policy exists and lives alongside the consent evidence in a single audit folder.

Is Krisp the answer when consent-less capture is unacceptable?

Krisp is the reference architecture when no meeting bot may join the call. The Krisp product page confirms audio processing runs on-device with cloud opt-in only for summaries; noise cancellation runs locally. The trade is thinner integrations and no cross-meeting reasoning. A governance-first team uses Krisp for external calls and a bot-based tool inside the org where consent is documented at the workspace level.

Are Gong and Chorus substitutes for meeting notetakers?

No. Gong Revenue AI and Chorus by ZoomInfo are sales-side conversation intelligence — deal intelligence, coaching, forecasting — and belong to a separate RFP. Do not conflate revenue-intelligence procurement with general meeting notes procurement. The overlap is technical (both capture calls); the buyer center and the pricing model are different.

The Brief The AI Layer series · piece 7 · Last updated July 2026

Meeting intelligence in 2026: when your bot's transcript becomes your institutional memory.

Q: Do we still need a third-party notetaker if we have Teams Copilot or Zoom AI Companion?

Not for baseline recap. Teams Copilot Video Recap reached GA April-May 2026 and recap-without-transcript reached GA early-to-mid June 2026 per Microsoft's Message Center notice. Zoom AI Companion 2.0 rolled out March 31, 2026 with meeting summaries, OCR on shared screens, side-panel AI, and AI templates. Third-party tools now have to justify incremental value — richer CRM sync, cross-meeting reasoning, or a specific integration the native tool does not ship.

Q: Where does meeting-derived knowledge belong?

One canonical destination, not several. Glean sits above Slack, Drive, Salesforce, Jira, and GitHub and crossed $200M ARR at a $7.2B valuation in early 2026. Guru offers AI knowledge management with 100+ integrations including Slack, Teams, Salesforce, Zendesk, Google Drive, and ChatGPT via MCP. Notion added Custom Agents on February 24, 2026 with AI Meeting Notes already in the workspace. Coda is a doc-app hybrid with database automation via Packs. Pick one destination and disable third-party bot exports to any other.

Meeting AI is no longer a productivity tool. It is a knowledge-ingestion pipeline most mid-market operators are running without a governance layer. Native recorders now cover baseline recap. Third-party notetakers earn their spend on richer CRM sync and knowledge-base ingestion. And Otter.ai's federal consolidated lawsuits make auto-join a legal liability in twelve all-party-consent states.

8 min · The AI Layer · Series piece 7

Questions this article answers

Which meeting-intelligence products dominate which segment (individual, sales, enterprise, security-sensitive)?
How do Teams Copilot and Zoom AI Companion 2.0 change the make-or-buy calculus?
Which knowledge platforms ingest meeting output natively?
What is the current two-party-consent and PII risk?
What retention posture should a mid-market operator set?
What is the governance checklist that keeps the pipeline defensible?

Meeting intelligence is no longer a productivity tool. It is a knowledge-ingestion pipeline that most mid-market operators are running without any governance layer. In 2026 the market bifurcated — native recorders inside existing UCaaS spend on one side, third-party notetakers pushing structured outputs into a knowledge base on the other — and the buyer's job is to write the policy that governs both.

The native shift is real. Microsoft Teams Copilot Video Recap rolled to GA April-May 2026, and recap-without-transcript reached GA early-to-mid June 2026 for M365 Copilot Premium. Zoom AI Companion 2.0 rolled out March 31, 2026 with meeting summaries, OCR on shared screens, side-panel AI, and AI templates. Baseline recap now ships inside the UCaaS license the buyer already pays for. Third-party notetakers — Otter, Fireflies, Fathom, Read.ai, Krisp, Grain, Circleback, tl;dv — have to justify incremental value against that new baseline: richer CRM sync, cross-meeting reasoning, a specific integration, or on-device processing that avoids a bot in the call.

The consent and litigation posture

Otter.ai faces a federal class action consolidated by Judge Eumi K. Lee into In re Otter.AI Privacy Litigation. Plaintiffs allege the AI Notetaker recorded meetings without all-party consent — consent obtained at most from the meeting host who added the bot, but not the other participants whose voices were captured. Motion to dismiss heard May 20, 2026 with no ruling as of publish. The tl;dv rundown tracks the four consolidated cases. The Recording Law state map lists twelve all-party-consent states — California, Connecticut, Delaware, Florida, Illinois, Maryland, Massachusetts, Montana, New Hampshire, Oregon, Pennsylvania, Washington — covering roughly 35 percent of the U.S. population. Per the Basil AI cross-state guide, the strictest applicable law governs a multi-state meeting. The immediate exposure sits with defaults that auto-join forwarded calendar invites; that is the fact pattern the consolidated complaint tests.

Which vendor for which buyer profile

Individual and small-team. Otter.ai covers up to three concurrent virtual meetings with Salesforce and HubSpot sync via Otter Sales Agent; Enterprise adds SSO and SOC 2. Fathom offers a free tier with unlimited recording (5 advanced AI calls per month cap); Premium at $19/month. Circleback ships single-meeting auto-action-item extraction.

Multi-platform capture and CRM automation. Fireflies.ai captures across Zoom, Meet, Teams, and wider platforms; 100+ integrations; 200+ AI Skills automations; starting at $10/user/month annual. tl;dv covers 30+ languages with timestamped highlights and AI agents for recording, note-taking, and CRM logging.

Meeting-behavior analytics. Read.ai adds participation and sentiment analytics. Grain compiles clips into Stories with HubSpot integration for sales and CS.

Consent-sensitive workflows. Krisp is the only major AI note-taker that does not require a bot to join the call; audio processing runs on-device with cloud opt-in for summaries. Pro tier at $8/month unlimited.

Native platform recorders. Teams Copilot and Zoom AI Companion 2.0 are the baseline; third-party tools must beat them on something specific.

Buyer-side. Supplier-paid. Buyers pay zero. Compensation has zero weight in the Cardinal Index scoring inside the Cardinal Method. Specific commercial terms live only in the private Decision Memo a buyer signs, never on this page.

Sales-side conversation intelligence is a separate procurement

Gong and Chorus by ZoomInfo capture calls too, but they belong to a different RFP. Gong is a revenue AI platform with deal intelligence and forecasting; Chorus adds ZoomInfo B2B enrichment and coaching. The buyer center is revenue leadership, not the CIO, and the pricing model is per rep, not per user. Do not conflate the two procurements — the overlap is technical, and the substitution risk is real if the enterprise treats a meeting notetaker as a Gong replacement.

Where meeting output belongs: one canonical destination

The knowledge-base decision is where the actual value compounds. Glean sits above Slack, Drive, Salesforce, Jira, and GitHub; Glean crossed $200M ARR at a $7.2B valuation in early 2026 on the Series F (cross-check the Glean press release before quoting the number precisely). Guru offers AI knowledge management with 100+ integrations including Slack, Teams, Salesforce, Zendesk, Google Drive, and ChatGPT via MCP; 2026 additions include Workday, Coda, and Miro. Notion added Custom Agents on February 24, 2026 with AI Meeting Notes already in the workspace. Coda is a doc-app hybrid with database automation via Packs, positioned as cheaper for large teams than Notion. Pick one canonical destination for meeting-derived knowledge, disable third-party bot exports to any other, and put the retention and attribution policy alongside the choice.

The buyer artifact

Meeting-to-knowledge governance checklist

Inventory every capture point. Native UC recorder, third-party bot, on-device tool, mobile side-channel. Label the lawful basis under the two-party consent map.
Set retention windows by artifact. Audio 30 days or less, transcript 90 days or less, summary 12 months or less. Override only with explicit business justification.
Pick one canonical knowledge destination. Glean, Guru, Notion, or Coda. Disable third-party bot exports to any other destination.
Require in-meeting notice and verbal acknowledgment for any recording that will be transcribed by AI, regardless of state. Defensive posture across the twelve all-party-consent states.
Configure the bot to auto-decline external meetings unless the organizer opts in. This defends against the Otter.ai fact pattern of a bot joining a meeting without participant consent.
Redact PII before ingestion. Names, account numbers, health data. Use vendor-native redaction (Otter, Fireflies) or a middleware layer.
Publish a Meeting AI Acceptable Use policy. Name permitted tools; prohibit recording of interviews, terminations, legal, and compliance meetings.
Quarterly audit. Sample twenty transcripts and verify consent evidence and redaction on each.

Where these deployments break

Failure mode one. Auto-join by default. Otter's default of joining forwarded calendar invites creates the exact fact pattern the consolidated complaint tests. Turn it off at the workspace level, or accept the vicarious risk when a Washington-based counterparty joins a California call.

Failure mode two. OCR on shared screens. Zoom AI Companion 2.0's OCR captures whatever appears on-screen, including PII in shared tabs or CRM records never intended for the transcript. Cover the redaction policy at the workspace level.

Failure mode three. Post-meeting automations that run without review. Fireflies' AI Skills can auto-email a summary; a mis-tagged meeting can send information to the wrong recipients. Require review-before-send on any automation with an external send.

Failure mode four. Native recap outputs skipping DLP. Teams Copilot and Zoom AI Companion recap can land in a personal OneDrive or Zoom cloud before hitting the governed knowledge base. Enforce the DLP surface at the workspace-recording setting.

Three rules for a mid-market operator

Rule one. Governance before ingestion. A retention and consent policy has to exist before the pipeline runs, not after it fails an audit.

Rule two. One canonical knowledge destination. Two destinations is what "knowledge sprawl" looks like in a governed environment; three is unrecoverable.

Rule three. Krisp is the reference architecture when consent-less capture is unacceptable. Use it where the counterparty would not consent to a bot, and route the rest of the workload through a governed native or third-party tool.

In short

Native meeting AI (Teams Copilot Video Recap GA April-May 2026, Zoom AI Companion 2.0 March 31, 2026) covers baseline recap. Third-party tools must justify incremental value.
Otter.ai faces federal consolidated litigation over consent-less capture; auto-join defaults are the exposure. Twelve all-party-consent states cover roughly 35 percent of the U.S. population.
Meeting output is a knowledge-ingestion pipeline. Glean, Guru, Notion, and Coda ingest transcripts differently. Pick one canonical destination.
Set retention by artifact — audio, transcript, summary — with override only on explicit business justification. Krisp is the reference architecture when no bot may join.
Gong and Chorus are separate procurements — sales-side conversation intelligence, not general meeting notes.
Publish an Acceptable Use policy that names permitted tools and prohibits recording of interviews, terminations, legal, and compliance meetings.

Sources

Microsoft 365 Message Center, “Video recap for Teams meetings (MC1261588),” April-May 2026 rollout. mc.merill.net
Microsoft Support, “Use Copilot without transcribing or recording a Teams meeting or call.” support.microsoft.com
USC ITS, “Zoom AI Companion 2.0 New Features,” March 31, 2026. itservices.usc.edu
tl;dv, “AI Meeting Recorder Lawsuits 2026.” tldv.io
Recording Law, “AI Meeting Recording Laws by State (2026).” recordinglaw.com
Basil AI, “Two-Party Consent AI Notetaker Compliance Guide 2026.” basilai.app
Otter.ai, pricing page. otter.ai
Fireflies.ai, “Fireflies vs. Fathom (2026)” comparison. fireflies.ai
Krisp, AI Note Taker product page. krisp.ai
Glean, “Comparing costs scaling AI search solutions in 2026.” glean.com

All linked sources were live at time of publish (July 2026). Verify before quoting in a procurement document.

Want a calibration on your meeting-AI governance?

Run the Tier 1 benchmark.

Submit your current UC contract, notetaker inventory, and knowledge-base list. We return a benchmark PDF in five business days mapping every capture point to a lawful basis, flagging duplicated ingestion into multiple knowledge destinations, and identifying which auto-join defaults expose the org under two-party-consent state law. Free. No follow-up sales drip.

Run the Tier 1 benchmark →

Series · The AI Layer

Read the anchor, Where AI actually fits in your tech stack. The UCaaS baseline at What your UCaaS already ships with. The conversational-AI disaggregation at Conversational AI taxonomy for the enterprise. And the Method framing at the Cardinal Method.