Microsoft 365 E7 adds tools to treat AI agents as insider risks
Representative image: succo / Pixabay
Microsoft said on October 19, 2026 it is adding controls to Microsoft 365 E7 and Project Perception that score AI agents' risk and can limit, pause or shut them down.
What Happened
Microsoft announced on Monday, October 19, 2026 a set of new security controls for businesses that use AI agents, the software programs that can carry out tasks on their own inside company systems. In a post on its security blog titled "Agents are your newest insiders", the company said it now treats such agents as potential insider threats, much like an employee with access to sensitive data. The new features, part of Microsoft 365 E7 and a security system called Project Perception, assess each agent's risk and can restrict what it does, ask for a human to step in, or shut it down.
Key Facts
- Announcement date: October 19, 2026, on Microsoft's security blog.
- The tools sit in Microsoft 365 E7, Microsoft's top business software bundle, and Project Perception, a security system Microsoft introduced in July 2026.
- Dynamic Permissions: according to Microsoft, agents start with no access and receive only the permissions needed for a specific task.
- Dynamic Agent Risk scores each agent using signals from identity, data, device, cloud and network activity.
- Dynamic Conditional Access steps up the response as an agent's risk score rises, from limits to human approval to shutdown.
- "Guardian agents" supervise other agents' actions continuously, the company said.
- Microsoft said its discovery tools now recognise 35 more types of locally run agents, 60 in all, and that Entra Agent ID will give agents identities separate from the people who set them up.
Why It Matters
Companies are moving fast from AI chatbots that answer questions to AI agents that act: sending emails, changing records, moving files or running code. That shift brings a new business risk. An agent works with real permissions inside company systems, so if it makes a mistake, is tricked by a malicious instruction hidden in a document or web page, or is taken over by an attacker, it can do damage at machine speed. Microsoft's guidance is to assume that an agent may fail and to limit its authority in advance.
The approach borrows from how firms already manage human staff. "Least privilege" means giving each person only the access they need for their job; "conditional access" means checking risk before allowing a login or action, for example from an unknown device. Microsoft is applying the same ideas to software agents, with the difference that the controls change automatically as the agent's behaviour changes. A separate identity for each agent matters because, without it, an agent's actions can look like those of the employee who set it up, which makes audits and investigations harder.
For business leaders, the announcement is also a sign of where costs are heading. Security for AI is becoming a paid layer on top of AI itself, bundled into premium licences. Companies weighing AI agents now have to budget not just for the AI tools but for governance: monitoring, identity management and incident response. The features are Microsoft's own description of what its products do; how well they work will depend on real deployments.
For Indian companies, many of which run on Microsoft's business software, the same questions apply. Firms in banking, insurance, IT services and other regulated sectors will need to show regulators and clients who authorised an agent's actions and how it was controlled.
| Control | What Microsoft says it does |
|---|---|
| Dynamic Permissions | Starts agents with zero access, then grants task-specific rights |
| Dynamic Agent Risk | Scores risk from identity, data, device, cloud and network signals |
| Dynamic Conditional Access | Escalates from limits to human approval to shutdown |
| Guardian agents | Continuously supervise other agents' actions |
| Entra Agent ID | Gives each agent its own identity |
Impact
Short-term: IT and security teams at companies on Microsoft 365 E7 can begin cataloguing the AI agents running on their systems and setting rules for them. Firms not on the top licence will have to decide whether the controls justify an upgrade.
Long-term: Treating AI agents like staff, with identities, permissions and supervision, is likely to become standard practice in enterprise IT. Agent governance may become a regular line in corporate technology budgets and in regulators' expectations.
Who is affected: Chief information officers and security heads, companies deploying AI agents, software vendors and IT service firms that build or manage agents for clients, auditors and regulators, and employees whose work is shared with automated agents.
Key Takeaway
Microsoft's October 19, 2026 update treats AI agents as insiders, giving businesses tools to score each agent's risk and to limit, supervise or stop it, a sign that AI governance is becoming a core business cost.
Questions and Answers
What did Microsoft announce for AI agents security on October 19, 2026?
Microsoft announced new controls in Microsoft 365 E7 and Project Perception that treat AI agents as potential insider threats. They score each agent's risk and can restrict it, require human approval or shut it down.
What is an AI agent?
An AI agent is software that uses artificial intelligence to carry out tasks on its own, such as sending emails, updating records or running code, rather than only answering questions like a chatbot.
Why does Microsoft call AI agents insiders?
Agents work inside company systems with real access to data and tools. Microsoft says they can make mistakes, be manipulated or be compromised, so they should be supervised and limited in the same way as employees with access to sensitive information.
What is Project Perception?
Project Perception is a Microsoft security system introduced in July 2026 to defend against AI-driven attacks. The October 19 update extends it to monitoring AI agents already running inside organisations.
Disclaimer: Prepared by the Peepals newsroom from publicly available sources with AI assistance. Information is accurate to the best of our knowledge at the time of publication and may change. Images may be representative. Not professional advice. Report an error via our contact page.
Sourced and fact-checked by the Peepals Global Editorial Team
Reported, fact-checked and published by the Peepals Global Editorial Team.
