Microsoft unveiled an AI Transparency Initiative at its Ignite conference, partnering with more than 40 companies—including Accenture, Adobe, KPMG, Mastercard and Volvo Cars—to craft open, industrywide standards for governing AI models and agents. The push, which builds on work with NIST, targets a growing blind spot for enterprises as AI agents operate with increased autonomy, often beyond current monitoring and audit controls. Microsoft plans to embed the standards into Microsoft AI, Foundry and Copilot so customers can generate verifiable evidence of model behavior for use with their existing security and compliance tools. Executives cast the effort as essential to accountability, arguing that “trust the model” is not a control. The approach centers on “see, verify, control” principles and is designed to be model- and provider-agnostic so companies can swap models without losing oversight. The move highlights intensifying pressure on vendors to make AI deployments safer and more auditable across regulated industries.
Related articles:
NIST AI Risk Management Framework (AI RMF 1.0)
Constitutional AI: Harmlessness from AI Feedback




























