Skip to content
AI.info

The Pulse

Microsoft and partners start enterprise AI transparency initiative

Microsoft is working with more than 40 organizations on proposed open standards for AI transparency, security and governance. The company plans to incorporate the work into Microsoft AI, Foundry and Copilot.

Microsoft and partners start enterprise AI transparency initiative

Standards across providers

Microsoft and more than 40 organizations are working on open standards intended to give enterprises greater visibility into AI models and agents. Chairman and CEO Satya Nadella announced the AI Transparency Initiative at Microsoft Ignite on Nov. 17, according to the company’s release.

The company says existing controls can observe much of the activity around models but cannot see what happens within them or independently verify their accounts of their own behavior. It describes that gap as a source of security, compliance, audit and accountability risk as agents gain access to enterprise data, tools and credentials.

Three parts to the plan

Microsoft will expand its participation in the National Institute of Standards and Technology’s SI Agent Standards Initiative and draw on its research to develop industrywide standards for AI transparency, security and governance. It will also work with organizations including Accenture, Adobe, KPMG, Mastercard and Volvo Cars on open standards across models and providers.

Microsoft plans to incorporate the work into Microsoft AI, Microsoft Foundry and Microsoft Copilot. The company says the resulting evidence will be usable with customers’ security and compliance tools, including those from other providers.

The initiative is designed to work with both closed and open-weight models. Microsoft says that design would let developers change or update the models behind agents without compromising security teams’ ability to govern them or compliance teams’ ability to audit their actions.

Seeing and verifying activity

Adobe Chief Security Officer Aanchal Gupta said security teams need visibility into what agents access, the actions they take and how those actions can be verified. She said a common industry approach is important for security and accountability.

Microsoft describes the desired approach as allowing organizations to see, verify and control AI activity. For visibility, it says organizations should know what an AI system was asked and authorized to do, which model and runtime it used, what data and tools it accessed, and what actions it took. Agent activity and risk signals should be monitored against policies and thresholds.

For verification, Microsoft says evidence about an agent’s identity, delegated intent and actions should identify its source. That could include the model, the runtime around it, or independent systems such as identity and policy systems, allowing model-reported signals to be checked against independently observed records.

Organizations should be able to check AI activity with their own security and compliance systems without relying solely on a model provider’s account, the company says.

Control and participation

The company also wants organizations to retain control over the data agents can access and how that access is governed, regardless of the model or provider. It says evidence about a workflow’s purpose, owner, data and delegated authority can help organizations enforce policies and reconstruct events during investigations and audits.

KPMG International Global Chief Information Security Officer John Israel said trust in AI requires transparency, explainability and responsible oversight. Organizations interested in helping shape the initiative or following its progress can sign up through Microsoft’s initiative page.

Source

  1. Microsoft Source

Companies in this story

Related stories

Explore

More articles