Darktrace among first cybersecurity firms selected to integrate risk signals into Microsoft Agent 365

Darktrace is among the first cybersecurity companies selected by Microsoft to contribute organisation-specific behavioural risk signals to the Microsoft Agent 365 portal.

Through a new integration, Darktrace / SECURE AI will derive these signals from each customer’s Unique Behavioural Profile and surface them directly within Agent 365. This will help customers understand which agents exist, what they are permitted to do and whether their actual behaviour is consistent with how that organisation normally operates.

As organisations deploy AI agents across business applications and workflows, those agents are gaining access to data, systems and tools and taking action on behalf of users. Their identities and permissions determine what they are allowed to do but security teams also need to understand how they behave.

The UK AI Security Institute (AISI) – which equip governments with a scientific understanding of the risks posed by advanced AI – recently published an incident report which shows autonomous AI agents independently developing novel attack chains, being able to adapt when detected and targeting humans and other AI systems.

How the new integration extends visibility within Agent 365

Microsoft Agent 365 provides a centralised way to manage and secure AI agents. Through the integration, Darktrace / SECURE AI will bring organisation-specific, behaviour-based agent risk signals into Agent 365 that reflect how each organisation’s agents normally operate and where their behaviour deviates from expectations. Customers will be able to view Darktrace-reported risk signals alongside Microsoft signals in a single interface.

Darktrace / SECURE AI is part of the Darktrace Behavioural Defence Platform, which provides unified visibility, continuous behavioural monitoring and autonomous response across AI, people and infrastructure. Powered by Darktrace’s Adaptive AI, it learns how an organisation operates and helps identify agent activity that may indicate compromise, manipulation, misconfiguration, or drift from intended behaviour.

Why it is wise to gain unified visibility of AI agent risk signals

Ed Jennings, President and CEO of Darktrace, said: “As AI adoption accelerates across the enterprise, agents are taking on more responsibility, accessing more sensitive data and acting across more systems.

“That makes it critical for security teams to understand not only which agents exist and what permissions they have, but how they actually behave over time. By bringing unique behavioural risk signals from Darktrace / SECURE AI into Microsoft Agent 365, customers can more easily identify agents that may require attention and investigate the deeper context behind the risk. This helps close the gap between what an agent is allowed to do and what it is actually doing, giving organisations greater confidence to scale AI adoption securely.”

Darktrace monitors how agents operate over time and how their actions connect to people, data and infrastructure, helping security teams identify unexpected interactions, behavioural changes and activity outside an agent’s intended purpose.

Nirav Shah, CVP, Agent 365 and Agent Cloud, Microsoft, said: “Microsoft is committed to delivering the industry’s most trusted and comprehensive AI governance and security experience as part of the Agent 365 promise. Through integrations with our security ecosystem partners, we enable customers to maximise the value of their existing security investments while gaining a unified, end-to-end view of risk signals across their agent ecosystem.”