<- All intelligenceThis Quarter
CISO
CTO
Enterprise Architect
Industry relevance
Financial Services
Healthcare
Government
OCTOBER 19, 2026
Microsoft Calls AI Agents Insiders and Introduces Task-Scoped, Time-Bound Agent Permissions, October 2026
Microsoft now treats agents as insiders and plans permissions limited to one task and one time window. The open question is who approves each scope and where that approval is recorded.
On October 19, 2026, Microsoft published a post by Nadim Abdo, Corporate Vice President, Identity and Network Access, that treats AI agents already inside the enterprise as insiders. Microsoft says it is expanding discovery to 35 additional types of local agents, for a total of 60 supported local agents, including Claude Code, OpenAI Codex, and GitHub Copilot. Microsoft says agents should begin with no authority and describes Dynamic Permissions scoped to the task and time-bound, with organization-wide policy setting the outer boundary. It also describes Dynamic Agent Risk, Dynamic Conditional Access, and Guardian Agents in Project Perception. The post says people define policy, approval boundaries, and delegation rules. It gives no availability dates for any capability.
GOVERNANCE IMPLICATION
Microsoft's post says agents should begin with no authority and describes permissions scoped to the task and time-bound, with organization-wide policy setting the outer boundary, including preventing an agent from obtaining authority its user does not have. That moves the accountability question upstream. A runtime log shows what an agent did and when its access ended. The harder record is who defined the delegation rules and approval boundaries the system then enforced, which the post assigns to people. The post gives no availability dates for the capabilities, so which parts exist in a given tenant today is a question for testing, not for the announcement.
SCENARIO
Hypothetical: A bank deploys an agent to reconcile vendor payments under task-scoped permissions. Ninety days later an examiner asks who approved the scope the agent held on a specific Tuesday. The runtime evidence shows the actions and the moment the access ended. The approval trail for the scope definition points to a policy last edited by an administrator, with no named business owner for the decision.
THE GOVERNANCE QUESTION
When Dynamic Permissions grants an agent authority scoped to one task, who approved that scope, and where does an auditor find the record of that approval?
CONTROL GAP
The post says people define policy, approval boundaries, and delegation rules, and describes runtime enforcement of them. Whether an organization's own process records who approved a given task scope, under which outer boundary, is something each tenant has to test.
REGULATORY RELEVANCE
NIST Ai RMF
PRIMARY SOURCE
Agents are your newest insiders. Security must move at their speed.
Nadim Abdo, Microsoft
October 19, 2026
Read the primary source ->(opens in new tab)CONTINUE READING
Read the next intelligence note.
JUNE 2, 2026
AgentsMicrosoft 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.
MARCH 24, 2026
AccountabilityAgent 365 GA governs on-behalf-of agents only — autonomous AI teammates with independent identities remain in Frontier preview beyond May 1
The May 1 GA version of Agent 365 governs agents operating on behalf of licensed human users. Agents with their own mailboxes, OneDrives, and independent identities map to the Agent Identity Authentication flow, which stays in Frontier preview. Frontier trial licenses were extended through December 2026, signaling that autonomous identity capability is still in development.
FEBRUARY 19, 2026
Identity DataMicrosoft Entra Agent ID governance requires a named human sponsor for every agent identity throughout its lifecycle to prevent orphaned agent IDs
Entra Agent ID governance establishes that every agent identity must have a designated sponsor — a human owner accountable for the agent's access and lifecycle. If the sponsor leaves the organization, sponsorship transfers automatically to their manager. Lifecycle workflows notify cosponsors and managers of impending sponsorship changes.