If an AI agent changes a customer’s access permissions, can the MSP show who authorized the change and what the agent actually did?
Microsoft launched its AI Transparency Initiative at Ignite 2026, bringing together more than 40 organizations to develop open standards for monitoring, verifying and governing AI models and agents across different providers.
Announced by Chairman and CEO Satya Nadella on Nov. 17, the initiative includes Accenture, Adobe, KPMG, Mastercard and Volvo Cars. The effort addresses a growing enterprise security problem: Organizations can monitor what AI agents access and the actions they perform, but often cannot independently establish what happens inside the models between a request and its outcome.
That creates potential security, compliance and audit problems as businesses give agents access to sensitive information, credentials, applications and infrastructure.
“For security controls to be effective, every organization needs greater transparency into the model itself,” said Hayete Gallot, Microsoft’s executive vice president of security.
How Microsoft’s transparency framework works
Microsoft is organizing the initiative around three capabilities:
| Capability | What it means for enterprises |
| See | Track what agents were instructed and authorized to do, which models they used, what information they accessed and what actions they performed. |
| Verify | Independently check agent activity against records from identity systems, runtimes and security tools rather than relying exclusively on a model’s explanation. |
| Control | Enforce permissions and organizational policies before agents perform consequential actions. |
Microsoft will expand its participation in the National Institute of Standards and Technology’s (NIST) SI Agent Standards Initiative and use its research to help develop industrywide standards.
It also plans to incorporate the resulting work into Microsoft AI, Microsoft Foundry and Copilot. The approach is designed to support both closed and open-weight models, allowing companies to change providers without losing their governance and audit capabilities.
Microsoft’s bigger bet on agentic security
The transparency announcement is part of a wider security push at Ignite.
Microsoft Perception, its AI-powered security orchestration platform, is now generally available. It coordinates red, blue and green team agents to simulate attacks, investigate threats and remediate vulnerabilities.
The company also introduced additional capabilities for discovering and governing enterprise agents through Microsoft Agent 365 and Microsoft Entra. These announcements address different parts of the same problem: knowing which agents exist, restricting their access and understanding their behavior.
What this means for IT teams and MSPs
For enterprise IT departments and managed service providers (MSPs), the initiative could eventually make AI governance less dependent on individual model providers.
That matters when organizations operate agents across multiple cloud platforms, applications and AI vendors. MSPs managing these environments could use independently verifiable activity records to support security investigations, compliance reporting and customer accountability.
However, the initiative is still developing. Microsoft has announced its standards ambitions and product integration plans, not a completed industry-wide verification system. Businesses should not assume existing agents automatically gain these protections.
The bigger challenge will be getting competing AI providers and security vendors to adopt compatible standards. If Microsoft and its partners succeed, enterprises could gain something increasingly important as AI adoption accelerates: the freedom to deploy agents without surrendering visibility and control over what they do.
A few changes to your prompts can make AI’s answers more useful. Learn how in How to Talk to AI, a free six-minute course from The Neuron Academy. To keep learning, explore the full course catalog free for seven days →


