MicrosoftAlym Rayani
Microsoft Adds Network Controls to Block Sensitive Data From AI Agents
Microsoft’s September 2026 security updates focus on discovering and controlling AI agents, protecting data shared with AI services and improving compliance operations. Microsoft Security Copilot users with Microsoft Defender can now access AI-generated “detonation summaries” explaining URL and file sandbox results in email investigations, reducing the need to manually correlate evidence and related signals.
Microsoft says Microsoft Purview and Microsoft Entra Global Secure Access are now generally available for enforcing data-security policies at the network layer, including traffic generated by agents acting on a user’s behalf. Organisations can identify sensitive files or text in real time and block uploads to risky, unsanctioned AI tools before data leaves the network.
Purview auto-labelling has also been expanded: simulations can cover up to 20 million items and 50,000 sites through adaptive scopes, while administrators can edit policies without rerunning simulations and access new audit and coverage reports.
Other updates include Purview eDiscovery support for searching, placing holds on, reviewing and exporting content in user-owned SharePoint embedded containers, including Microsoft Loop, Copilot Pages and Copilot Notebooks. Administrators can archive inactive SharePoint content while retaining it for legal discovery and retention requirements, removing it from Microsoft 365 Copilot indexing until reactivated. Approved stale content, such as Teams recordings and transcripts, can also be permanently deleted.
Microsoft further announced that Intune Enterprise Application Management, Cloud PKI and Remote Help are coming to GCC High, with Enterprise Application Management also available to Department of Defense organisations.