Zoom separates available sensitivity labels from planned AI governance tools

Zoom’s [October 26 governance update](https://www.zoom.com/en/blog/zoomtopia2026‑ai‑governance‑zoom‑workplace/) separates a control available now from a set of AI features planned for the coming months. Sensitivity Labels are available for Zoom Meetings and Whiteboard; Centralized AI Management, custom AI guardrails, Deepfake risk detection and expanded Compliance Manager coverage have later expected dates. The post explains announcements made at Zoomtopia, rather than saying all five capabilities launched on October 26. Organizations can create labels within Zoom or optionally sync them from Microsoft Purview. Purview is not a prerequisite for using native Zoom labels. Zoom says the labels are included in Enterprise Plus and Premier accounts and can be purchased as an add‑on for other licences. The announcement does not provide add‑on prices. For an administrator, the immediate task is classification: deciding how meeting and whiteboard information should be labelled. That is a different function from setting rules for what an AI agent may process or put into an answer.

[Centralized AI Management is expected in December 2026](https://www.zoom.com/en/blog/zoomtopia2026‑ai‑governance‑zoom‑workplace/). Zoom describes a single administrative view of agent activity, account‑level policies, audit controls and oversight of agents employees create. Those are planned service capabilities, not evidence that an organization already has a complete inventory or audit trail. Custom AI guardrails are expected in the first quarter of 2027. Zoom says administrators will be able to define rules about information AI can access and include in outputs. Its examples include excluding certain channels or meeting types from processing and requiring human review before particular content is shared. The distinction matters when planning a rollout: the classification labels available now are not a substitute for those later agent‑specific controls. Teams should match a proposed workflow to the control actually available at the time, rather than treating the roadmap as a completed configuration.

Zoom expects [Deepfake risk detection in December 2026 as a paid add‑on](https://www.zoom.com/en/blog/zoomtopia2026‑ai‑governance‑zoom‑workplace/). The company says it will analyse audio and video in meetings, flag content that may have been modified outside Zoom and display a confidence indicator. The planned output is a risk signal participants can investigate, rather than an identity certificate. Zoom explicitly says the feature will not catch everything and is not a substitute for other verification methods. The announcement supplies no detection‑accuracy or false‑positive results. Its examples of hiring interviews and customer interactions describe intended uses, not independently demonstrated effectiveness.

[Compliance Manager coverage for My Notes and Zoom Canvas is expected in January 2027](https://www.zoom.com/en/blog/zoomtopia2026‑ai‑governance‑zoom‑workplace/), extending retention and archiving controls to content in those products. Zoom distinguishes that planned expansion from the coverage Compliance Manager already provides for other communication and collaboration services. This addresses a separate administrative question from labelling or live AI output controls: what records are retained and archived after people and agents have produced them. Zoom’s schedule gives administrators a set of availability checkpoints; it does not establish that a particular deployment meets an organization’s record‑keeping or compliance obligations. *Photo: [Coolcaesar](https://commons.wikimedia.org/wiki/user:Coolcaesar) / [Wikimedia Commons](https://commons.wikimedia.org/wiki/File:Zoomheadquarters.jpg), [CC BY‑SA 4.0](https://creativecommons.org/licenses/by‑sa/4.0/). Zoom headquarters on Almaden Boulevard, San Jose, June 21, 2020. Archive company context, not a depiction of the governance tools or their deployment. Geometric crop and downscale; image adaptations retain the same licence.*