Newshot

Microsoft Joins Accenture, Adobe, Mastercard, Volvo and 40+ Other Organizations to Advance AI Standards, Stressing Observability, Verification and Control Across Both Closed and Open-Weight Models

AI summary

At Ignite, Microsoft announced the launch of an AI Transparency Initiative with more than 40 organizations, including Accenture, Adobe, KPMG, Mastercard and Volvo Cars, to jointly develop open standards applicable to different AI models and service providers in order to strengthen the transparency, security and governance of AI systems. Microsoft Chairman and CEO Satya Nadella said the company will work with customers and partners to establish an open, universal AI safety and governance framework, noting that as enterprises connect AI models to internal data, tools, credentials and infrastructure, AI agents can already execute tasks autonomously without step-by-step human approval, yet existing security management mechanisms typically capture only partial information about AI activity and cannot reconstruct the full process from request to result or independently verify a model's description of its own behavior. The initiative focuses on expanding participation in NIST's AI Agent Standards Initiative, working with the 40+ organizations on open standards that apply to both closed-source and open-weight models so enterprises can keep using existing IT security management and complian

Translation(model translation, not the source's own text)

Title: Microsoft Joins Forces with Accenture, Adobe, Mastercard, Volvo and 40+ Other Organizations to Advance AI Standards, Emphasizing Observability, Verification and Control, Covering Both Closed-Source and Open-Weight Models

Body:

Microsoft Joins Forces with Accenture, Adobe, Mastercard, Volvo and 40+ Other Organizations to Advance AI Standards: Emphasizing Observability, Verification and Control, Covering Both Closed-Source and Open-Weight Models

IT之家 November 18 news: Microsoft today at the Ignite conference announced the launch of the AI Transparency Initiative, working with more than 40 organizations including Accenture, Adobe, KPMG, Mastercard and Volvo Cars to jointly advance the development of open standards applicable to different AI models and service providers, in order to strengthen the transparency, security and governance of AI systems.

Microsoft Chairman and CEO Satya Nadella announced that the company will work with customers and partners to advance the establishment of an open, universal AI safety and governance framework.

Microsoft said that as enterprises connect AI models to internal data, tools, credentials and infrastructure, AI agents are already able to autonomously execute tasks without requiring step-by-step human approval. However, existing security management mechanisms typically can only observe partial information about AI activity, making it difficult to reconstruct the complete process from when a model receives a request to when it produces a result, and impossible to independently verify a model's description of its own behavior. This brings new challenges to enterprises' security management, compliance auditing and accountability tracing.

The initiative mainly consists of three aspects. First, Microsoft will expand its participation in the U.S. National Institute of Standards and Technology (NIST) AI Agent Standards Initiative and use related research to help advance the development of industry open standards covering AI transparency, security and governance.

Second, Microsoft will work with more than 40 organizations to advance the establishment of open standards applicable to different models and service providers, so that AI systems are accountable, independently verifiable, and incorporate security mechanisms at the design stage. Microsoft emphasized that this set of standards will apply to both closed-source models and open-weight models, so that when enterprises replace or upgrade their underlying models, they can continue to use their existing IT security management and compliance auditing mechanisms.

Third, Microsoft plans to integrate the related results into Microsoft AI, Microsoft Foundry and Microsoft Copilot products and services, enabling enterprise customers to connect the evidence generated by AI activity into security and compliance tools provided by Microsoft or other vendors.

In terms of concrete implementation, Microsoft summarizes the initiative into three stages: observe, verify and control. IT之家 summarizes as follows:

  • The observe stage requires enterprises to be able to understand what instructions AI received, what authorizations it obtained, what model and runtime environment it used, what data and tools it accessed, and what operations it actually executed, while continuously monitoring agent activity and risk signals.
  • The verify stage emphasizes that AI agents cannot be the sole source of information proving their own behavior. Enterprises need to cross-verify AI identity, authorization intent and actual operations by combining records provided by the model, the runtime environment, and independent systems such as identity authentication and policy management, rather than relying solely on the model service provider's description of its behavior.
  • The control stage requires enterprises to retain the ability to manage AI permissions, including restricting the data agents can access, the operations they can execute and the authorizations they obtain. When a security incident occurs, enterprises should also be able to use the relevant records to reconstruct the course of the incident and use them for investigation, risk control, recovery and auditing.

Hayete Gallot, Executive Vice President of Microsoft Security, said that for security controls to truly work, enterprises need greater transparency into what is happening inside models.

Adobe Chief Security Officer Aanchal Gupta and KPMG International Global Chief Information Security Officer John Israel also said respectively that enterprises need a clearer understanding of what resources AI agents have accessed, what operations they have executed, and how to independently verify these behaviors.

Microsoft said that as AI agents become further integrated into enterprise operations, organizations must continue to bear responsibility for their behavior and cannot use AI itself as a reason to evade responsibility.

Related Articles

  • Microsoft's in-house chip Maia 200 enters production: 1.7x improvement in token performance per watt
  • Microsoft's decision model Microsoft-Decision-1.1 debuts, outperforming GPT-6 Luna and Gemini
  • Microsoft and Apple outages in succession, Google responds urgently to Gmail outage rumors: "service operating normally"
  • Multilingual error rate drops 50%: Microsoft upgrades doctor-patient conversation AI transcription engine, now covering 1/3 of physicians in the U.S.
  • Microsoft's first speech-to-speech model: MAI Realtime 1 debuts, starting at $0.01 per minute
  • Microsoft's most cost-effective cybersecurity AI model MAI Cyber 2 Flash debuts: performance on par with GPT-6 Astra, cost per task about 2% of it

The translation is machine-generated and individual terms may be off. Nothing is rewritten or cut; the original is authoritative.

Published by IT之家
This site only summarises and aggregates; copyright stays with the original author. Headlines and summaries follow the source — where they differ, the original wins.
Read the original ↗