Microsoft improves AI bot defences in Teams

Microsoft is putting the brakes on artificial intelligence bots joining Teams meetings with improved detection and a new policy for access control in its flagship conferencing app.

By using what Microsoft says is a combination of behavioural and infrastructure signals, Teams' ability to distinguish between AI notetakers and humans is now stronger.

A new admin policy that is enabled by default also marks the bots as such, and lets Teams meeting organisers decide whether or not to let them join has also been introduced.

With the policy enabled, meeting participants are grouped as "waiting" which includes verified or regular attendees, and registered bots.

Unregistered or bots detected by Teams are categorised as "suspected threats" if the new policy is enabled, for quick identification.

Microsoft has also made harder to accidentally admit bots waiting in the Teams lobby to meetings, with confirmation prompts and warnings.

The change was brought in to address security and privacy risks from third-party AI bots joining meetings, capturing and transcribing these without other participants awareness.

Microsoft is also intending to set up a Teams bot identification program that lets software vendors register their note-takers and meeting assistants, and to clearly mark them as such.