WORLDTECH NEWS Global technology intelligence.Contact

Microsoft launches AI Transparency Initiative with more than 40 organisations

Monochrome image of a laptop, camera, lens, and coffee cup on a wooden desk
Illustrative photo.Photo by Pixabay on PexelsMicrosoft logo shown for identification only; no affiliation with or endorsement of WORLDTECH is implied.

Microsoft has launched the AI Transparency Initiative, a joint effort with customers and partners to build a shared, open framework for securing and governing AI models and agents. Chairman and CEO Satya Nadella presented it at Microsoft Ignite as the next stage of the company's work on safer AI.

What the initiative is

Nadella unveiled the initiative at Microsoft Ignite, describing it as the latest phase of Microsoft's commitment to making AI safer. It is structured as a collaboration with customers and partners, aiming at one open approach to security and governance that covers models from many providers, not just a single vendor's technology stack.

Microsoft frames the effort around a basic idea: the more autonomy AI is given, the more transparency it needs. The company points out that enterprises are linking AI models to data, tools, credentials and infrastructure, so agents can carry out actions without a person signing off on each step. Organisations, it argues, therefore need to be able to follow what those agents are doing.

Participants and the three-part plan

Microsoft says it is working with more than 40 organisations, among them Accenture, Adobe, KPMG, Mastercard and Volvo Cars, on open standards for AI systems that can be held to account, verified independently and designed for safety from the outset, whichever model or provider is involved.

The initiative has three components. Its stated goal is that every organisation can see what AI systems reach and do, check those actions independently and decide what the systems may do. Microsoft describes the seeing part as knowing what an AI was asked and permitted to do, which model and runtime it ran on, which data and tools it touched and what it actually did, with agent activity watched continuously against policies and limits.

The evidence gap Microsoft points to

Microsoft's case is that current controls can observe much of the activity around AI models but cannot see inside them or independently test what they report about their own conduct. Organisations often cannot piece together what happened between a request and its result, which weakens their grasp of why an agent acted as it did and raises risks for security, compliance, auditing and accountability.

The verification element is built on the premise that an agent should not be the only authority on its own behaviour. Records of agent identity, delegated intent and actions should show where they came from, whether the model, the runtime or independent systems such as identity and policy tools, so that signals reported by a model can be cross-checked against independently observed records.

Industry voices on the initiative

Aanchal Gupta, Chief Security Officer at Adobe, said security teams need clear visibility into what AI agents access, which actions they take and how those actions can be verified as agents become more deeply embedded in enterprise operations. Hayete Gallot, Executive Vice President of Microsoft Security, said that for security controls to work, every organisation needs more transparency into the model itself. Harpreet Sidhu, global cybersecurity lead at Accenture, said organisations taking the next step in their AI reinvention need a solid foundation of trust to build on.

Open questions

  • Which concrete open standards will the participating organisations publish, and on what timetable?
  • How will the approach be applied in practice to both closed and open-weight models?

Key facts

  • Microsoft Chairman and CEO Satya Nadella announced the AI Transparency Initiative at Microsoft Ignite on Tuesday.
  • The initiative is a collaborative effort with customers and partners to set a common, open approach to security and governance across models and providers.
  • Microsoft's underlying principle is that greater AI autonomy requires greater transparency.

Sources & evidence