microsoft logo in front of building

Microsoft Launches AI Transparency Initiative to Secure Enterprise Agents

Backed by more than 40 organizations, the open framework helps security teams track, verify and control autonomous AI activity across platforms.

Security teams now have a framework to monitor autonomous artificial intelligence after Microsoft launched the AI Transparency Initiative with more than 40 partners, including Accenture, Adobe, KPMG, Mastercard and Volvo Cars.

The effort aims to establish open, industrywide standards for AI security, governance and traceability across different models and providers.

Enterprises increasingly connect AI agents to internal data, software tools and login credentials, allowing them to execute tasks without step-by-step human approval. Existing security controls observe surface-level activity but often cannot reconstruct what happens inside a model between a request and an outcome. This visibility gap creates compliance, audit and insider-threat risks as agents operate at machine speed.

The initiative addresses these risks through three components:

  • Open standards development: Microsoft is expanding its work with the National Institute of Standards and Technology's AI Agent Standards Initiative to establish universal benchmarks for accountable, verifiable AI systems.
  • Cross-industry collaboration: The coalition will co-develop open governance frameworks that apply across closed and open-weight models, allowing developers to swap underlying models without breaking IT security controls.
  • Product integration: Microsoft will embed these visibility tools into Microsoft AI, Microsoft Foundry and Microsoft Copilot, allowing organizations to feed activity records directly into existing third-party security and compliance systems.

The framework centers on a see, verify and control architecture. Rather than relying on a model to accurately report its own behavior, the standard requires independent telemetry covering agent identity, delegated intent and data access. Security teams can cross-reference model-reported signals against external identity and policy logs to enforce guardrails before actions occur or to reconstruct events during an audit.

Industry security leaders from partners such as Adobe, KPMG and Accenture noted that as AI agents embed deeper into business operations, organizations require standardized visibility and independent verification to deploy the technology at scale without sacrificing compliance or control.

Organizations can follow the project's progress or register to participate through Microsoft's initiative portal.

About the Author

Jesse Jacobs is assistant editor of SecurityToday.com.

Featured

New Products

  • Product Image

    Teledyne FLIR's A40 and A70

    UL 2684-listed thermal cameras monitor temperature changes and provide independent testing evidence for European fire detection projects.

  • ROAMEO and SARA new product image

    AITX ROAMEO™ and SARA™

    Deploy a unified, self-coordinating outdoor security presence by combining ROAMEO’s autonomous patrol capabilities with SARA’s proactive event assessment and real-time response.

  • new product

    Adams Rite 4300 Steel Hawk® Electrified Deadlatch

    Modernize aging storefront doors with a refined electrified deadlatch engineered to overcome sag, misalignment and heavy wear without replacing existing hardware.