AI, in your work.

What Microsoft has announced

Microsoft Teams is preparing changes that would let organisers manage who can enable and use AI in their meetings. For IT, Microsoft announces a simpler policy model covering experiences during and after the session. The changes appear in its public roadmap digest.

Planned availability

The official entry 577399, created on 22 October 2026 and modified on 23 October, remains in development. It targets general availability in November 2026 for Worldwide (Standard Multi-Tenant), with Android, desktop, iOS, Mac and web among the listed platforms.

Microsoft warns that dates and descriptions on its roadmap are estimates subject to change. This forecast does not confirm that the controls are available today or guarantee an activation date for each organisation.

How control is currently divided

The current administrator documentation distinguishes four Copilot policy values, alongside their interaction with the transcription policy. These identifiers describe the configuration currently documented.

Licensing, transcription and access

Configuring AI use and holding a licence to use it are separate matters. Microsoft Support explains that organisers can adjust the options even without a Copilot licence. Participants using it need the appropriate Microsoft Copilot licence.

The same guide distinguishes use during the meeting, which can work without saving a transcript, from later questions about the spoken discussion, which require an available transcript. Turning off Copilot for the meeting also disables recording and transcription; Copilot in the chat can still answer questions about chat messages.

In its guidance on use without transcription, Microsoft warns that prompts and responses may be retained under Microsoft Purview policies, even without recording or transcription. It also states that external participants cannot use the meeting-only Copilot mode, even if they hold a licence.

Nexo's interpretation: what to validate

Our editorial interest is how responsibility for AI is assigned: which rules IT establishes and which decisions organisers make for individual sessions. We propose these hypothetical scenarios for evaluating the controls when they become available:

• Internal project review: check who can start AI, who can use it and which materials remain accessible afterwards.
• Client meeting: test internal and external access separately, without assuming that a licence from another organisation enables the same functions.
• Session involving sensitive information: compare recording, transcription, consultation permissions and retention with the rules agreed for that type of conversation.

What remains unspecified

The entry does not detail permissions by role, every affected AI experience, specific licensing requirements or the mapping to existing policies. It does not describe an automatic migration either. These are limits of the information published in entry 577399.

We recommend documenting current settings and preparing those checks before changing organisational rules. Assessment of the new model will need to draw on its rollout instructions and behaviour that can be verified when it arrives.