All AI news

Microsoft outlines new AI-agent security controls

Microsoft describes extending Project Perception to supervise internal agents and adjust their access as risk changes.

Announced 2026-10-19; timezone unspecified. Microsoft describes task-scoped Dynamic Permissions and adaptive risk/access controls. Editorial view: a useful prompt for teams to review excessive agent access to company data. Not tested; local activation dates and licensing remain unconfirmed.

Usage conditions

Hong Kong
Enterprise security announcement; tenant eligibility, plans, rollout and pricing for Hong Kong are unconfirmed. This does not establish consumer Copilot availability.
Macau
Enterprise security announcement; tenant eligibility, plans, rollout and pricing for Macau are unconfirmed. This does not establish consumer Copilot availability.
Taiwan
Enterprise security announcement; tenant eligibility, plans, rollout and pricing for Taiwan are unconfirmed. This does not establish consumer Copilot availability.

Official source: Agents are your newest insiders. Security must move at their speed.

Source published:

Source checked:

Published here: