Microsoft Rebuilds Its Responsible AI Standard Around the Technology Stack and Points Governance at Agent Identities and Tool Permissions, September 2026

Microsoft now structures its responsible AI rules by technology layer and names agent identities, tool permissions, and action monitoring as controls. Check the mapping to your inventory.

A compliance professional and technology risk analyst review policy and audit documents. — Compliance

On September 1, 2026, Microsoft published its third annual Responsible AI Transparency Report. Chief Responsible AI Officer Natasha Crampton wrote that Microsoft re-engineered its Responsible AI Standard so that it is structured by technology stack component, covering models, platform services, and applications, and by the role Microsoft plays in each. Microsoft says its technical risk management now focuses on controls such as agent identities, tool permissions, and monitoring of actions. Tools named include an AI Red Teaming Agent, agent evaluators, RAMPART, ASSERT, and the Agent Control Specification. Microsoft states it is one of the few companies certified against ISO 42001 across a broad portfolio.

GOVERNANCE IMPLICATION

The report describes a vendor whose internal governance moved to requirements organized by models, platform services, and applications, with a core set that always applies and scenario-specific requirements that change as capabilities and risks change. The control language that matters to customers is agent identities, tool permissions, and monitoring of actions, which Microsoft treats as central to governing systems that retain memory, use tools, and act for users. An enterprise inventory organized by business use case describes the same system from a different angle than the vendor's standard does. A failed control belongs to whoever owns the layer that produced it, and a use-case inventory does not say which layer that was.

SCENARIO

Hypothetical: A financial services firm's AI register lists agents by business use case. After a data exposure, the incident team has to decide whether the failing control belonged to the model, the platform service, or the application. The register records the use case and the sponsor, and cannot answer the layer question without a separate investigation.

THE GOVERNANCE QUESTION

Microsoft governs by technology layer. If your AI inventory is organized by use case, who owns the control when the two views disagree?

CONTROL GAP

Inventories organized by business use case do not record which technology layer each control sits in. The vendor's standard, certification scope, and published tooling are organized by layer, and the mapping between the two is work the customer has to do.

REGULATORY RELEVANCE

ISO 42001

PRIMARY SOURCE

Responsible AI in 2026: How we are adapting for what's ahead

Natasha Crampton, Microsoft

September 1, 2026

Read the primary source ->

Read the next intelligence note.

Back to Compliance

An IT administrator and a business content owner review access together at a laptop.

JUNE 2, 2026

Agents

Microsoft Agent Control Specification Brings Deterministic Runtime Enforcement, Leaving the Authorization Decision Upstream, June 2026

On June 2, 2026, Microsoft announced the Agent Control Specification (ACS) and ASSERT at Build 2026, authored by Sarah Bird on the Microsoft Foundry Blog. ACS is an open industry specification, part of the Agent Governance Toolkit, that places deterministic safety and security controls at five validation checkpoints in an agent's lifecycle: input, LLM, state, tool execution, and output. Controls are expressed as portable, versionable, auditable policy and are designed to work across any agent framework. ASSERT, a separate open-source project, converts written policies into executable evaluation scenarios. ACS launched with customer and partner endorsement including KPMG, Zscaler, IBM, and Arize AI.

Read note ->

<- Back to all intelligence notes