Collecting Microsoft 365 Copilot Data with Microsoft Purview eDiscovery
Copilot Data Collection Reference Table Data Type Storage Location Item Class Collection Strategy Copilot Prompts (user questions sent to M365 Copilot) Exchange Online: Hidden folder in the user's mailbox. Compliance copies stored similar to Teams chats, but with unique item classes. IPM.SkypeTeams.Message.Copilot.<AppName> (e.g., .Word, .Excel, .Outlook, .BizChat). Additional AI-related classes may also apply: IPM.SkypeTeams.Message.ConnectedAIApp*, IPM.SkypeTeams.Message.CloudAIApp*, IPM.SkypeTeams.Message.TeamCopilot*, IPM.SkypeTeams.TeamCopilot* 1. Add the user's Exchange mailbox as a data source to the search. 2. In the condition builder you can optionally filter the search to only return Copilot prompts by adding a condition of "Item class contains any of Copilot activity". This automatically applies all relevant M365 Copilot item classes as a condition of the search. 3. Add any further additional conditions such as date range or keywords to narrow results as required. You can also use the Item Class condition to exclude M365 Copilot interactions from your collections when targeting a user’s mailbox. Notes: · Additional item classes may be added. The item class condition will be updated accordingly. Copilot Responses (AI-generated answers) Exchange Online: The same hidden folder in the user's mailbox as prompts. The same IPM.SkypeTeams.Message.Copilot.<AppName> pattern as prompts The same collection strategy used for prompts. Copilot Memories (personalized saved information Copilot "remembers") Exchange Online: Hidden CopilotMemory subfolder within the user's mailbox contacts. Stored as contact entries separate from prompts and responses. IPM.Contact Each memory item appears as a contact card within Exchange, which is distinct from the message-based item classes used for prompts/responses. 1. Add the user's Exchange mailbox as a data source to the search. 2. In the condition builder you can optionally filter the search to only return Contacts by adding a condition of "Item class contains any of Contacts". Notes: · Copilot memories will not be preserved under a legal hold or retention policy. · This will return both Copilot memories stored in contacts as well as traditional contacts from the user’s Exchange mailbox. Copilot Pages (AI-generated, user-editable documents) SharePoint Online: Stored in a user-owned SharePoint embedded container (shared with Loop workspace content and Copilot Notebooks). File format is .page. Not stored in the user's mailbox. N/A These are SharePoint files (not Exchange items), so no item class applies. Identify them in search results by the .page file extension. 1. Add the custodian’s SharePoint embedded site URL as a data source to the search. Alternatively, tenant-wide searches of all SPO sites will include all SharePoint Embedded containers 2. Optionally use the condition builder with conditions such as date range, keywords or file type to further filter results returned Facilitator agent interactions in a Team meeting chat Exchange Online: Hidden folder in all meeting attendees’ mailboxes. Compliance copies stored as Teams chats IPM.SkypeTeams.Message 1. Add the user's Exchange mailbox as a data source to the search. 2. In the condition builder you can optionally filter the search to only return Copilot prompts by adding a condition of "Item class contains any of Instant messages". 3. Add any further additional conditions such as date range or keywords to narrow results as required. Facilitator agent meeting notes (loop) SharePoint Online: Facilitator meeting notes are stored as a .loop file in a OneDrive folder titled Meetings of the user who initiated Facilitator in Teams N/A These are SharePoint files (not Exchange items), so no item class applies. Identify them in search results by the .loop file extension. 1. Add the user's OneDrive URL as a data source to the search. 2. In the condition builder you can optionally filter the search to only return loop files by adding a condition of "File type equals any of loop". 3. Add any further additional conditions such as date range or keywords to narrow results as required. Notes: · With eDiscovery premium enabled cases you can follow the standard workflow for collecting Team meeting messages and select to include cloud attachments in your collection. This will automatically pull into the export or review set any Facilitator agent meeting notes. Facilitator created word/loop documents SharePoint Online: When the facilitator agent is asked to create a word or loop document during a meeting they are stored in the requesters OneDrive in a folder called N/A These are SharePoint files (not Exchange items), so no item class applies. Identify them in search results by the .loop file extension. 1. Add the user's OneDrive URL as a data source to the search. 2. In the condition builder you can optionally filter the search to only return loop and doc files by adding a condition of "File type equals any of loop, docx". 3. Add any further additional conditions such as date range or keywords to narrow results as required. Notes: · With eDiscovery premium enabled cases you can follow the standard workflow for collecting Team meeting messages and select to include cloud attachments in your collection. This will automatically pull into the export or review set any Facilitator generated loop or word documents. Facilitator generated and assigned tasks Exchange Online: When the facilitator agent creates and assigns a task to an individual, it is created as a to-do item in the assigned individual's Exchange Mailbox IPM.Task 1. Add the user's Exchange mailbox as a data source to the search. 2. In the condition builder you can optionally filter the search to only return Tasks by adding a condition of "Item class contains any of Tasks". 3. Add any further additional conditions such as date range or keywords to narrow results as required. Application-Specific Item Classes for Prompts & Responses For more granular filtering by Copilot application, the following item class values can be used in KQL queries: Application Context Item Class Value Microsoft Copilot Chat (BizChat / Teams) IPM.SkypeTeams.Message.Copilot.BizChat Copilot in Excel IPM.SkypeTeams.Message.Copilot.Excel Copilot in Loop IPM.SkypeTeams.Message.Copilot.Loop Copilot in Outlook IPM.SkypeTeams.Message.Copilot.Outlook Copilot in PowerPoint IPM.SkypeTeams.Message.Copilot.PowerPoint Copilot in Teams IPM.SkypeTeams.Message.Copilot.Teams Copilot in Whiteboard IPM.SkypeTeams.Message.Copilot.Whiteboard Copilot in Word IPM.SkypeTeams.Message.Copilot.Word To target all Copilot applications at once, use the wildcard query ItemClass:IPM.SkypeTeams.Message.Copilot.*. For a wider list of AI data sources, see the following link: https://learn.microsoft.com/en-us/purview/edisc-search-copilot-data#data-sources-for-ai-data Important Notes for eDiscovery Practitioners Excluding Copilot Data from Broader Searches Because Copilot prompts and responses reside in the same Exchange mailbox as emails and Teams chats, they will appear in broad mailbox searches unless explicitly filtered out. To exclude Copilot items, use the condition "Item Class Contains none of Copilot activity" in the condition builder, or add (-ItemClass:IPM.SkypeTeams.Message.Copilot.*) in KQL. Some eDiscovery managers run separate searches, one for Copilot data and one for other communications, to keep collections distinct. Copilot Memories: Retention & Hold Limitations Purview retention policies and eDiscovery holds do not currently apply to Copilot memory items. Memory items remain until a user deletes them or an admin explicitly removes them via eDiscovery or Graph API. Additionally, deleting a Copilot prompt and response does not delete any memory derived from that conversation. Memories must be removed separately if required. Copilot Pages: Do Not Treat Like Prompts/Responses Copilot Pages are not stored in Exchange mailboxes. Searching only a custodian’s mailbox will not return Copilot Pages. Treat Copilot Pages the same way as you do for SharePoint content in your existing eDiscovery workflow. For collections, keyword searches will generate hits on text content within the .page file if either the SharePoint Embedded URL is included in the search or the search is a tenant-wide search of all SharePoint sites Be aware that full-text search within .page files in Purview eDiscovery review sets is not currently available. Instead you can use filters such as Subject/Title or Native File Type to locate Copilot Pages in your review set and review the content. When an eDiscovery hold is placed on a custodian’s mailbox, it does not automatically extend to the SharePoint Embedded site where the Copilot Pages are stored. Instead, ensure the hold policy includes the URL for the user-owned SharePoint Embedded site that contains the Copilot Page(s) that must be preserved. Audit Logs vs. eDiscovery for Copilot Content Audit logs record that a Copilot interaction occurred (time, user, workload context) but do not include the actual prompt or response text. To retrieve the substance of Copilot interactions, use Purview eDiscovery searches against the mailbox. Copilot Prompts and Responses: HTML Transcription Copilot prompts and responses are stored as individual messages within the user’s mailbox. When collecting Copilot interactions, enabling the “Organize conversations into HTML transcripts” premium option will convert these individual messages into HTML transcripts making for easier review and linkage between the user’s original prompt and the Copilot responses. Copilot Prompts and Responses: Contextual prompts and responses When using the Keywords condition as part of your collection in eDiscovery, it will only return items that match the keywords included in the query. This means that you may only return a part of the Copilot interaction. If using keywords in your collection query you can enable the “Include full conversation for Copilot, Teams and Viva Engage messages” premium option. This will include in the export or review set any prompts or responses from the Copilot interaction within a 12-hour window before and after each responsive item. This means that you are able to see the full context of the prompt or response that was responsive to search. Collecting Referenced Documents (Cloud Attachments) Copilot responses may reference or summarize SharePoint/OneDrive files. When collecting Copilot interactions, enabling the "Access links (cloud attachments) in messages" premium option will additionally collect the files referenced in the prompt or response and include them in the export package. This provides full evidentiary context but can significantly increase export size and processing time so consider if collecting these artifacts are relevant to the investigation. If so, look to use additional conditions such as date to effectively manage volumes or reduce the number of custodians in the collection. Facilitator agent in Microsoft Teams Meetings The Facilitator agent in Microsoft Teams is an AI-powered assistant (included with Microsoft 365 Copilot) that enhances meeting productivity by generating real-time notes, summarizing key decisions, and managing action items. It acts as an active participant, allowing for collaborative editing of notes and answering chat questions during calls. As the Facilitator works within the context of Microsoft Teams meetings (scheduled private meetings only) your existing workflows for collecting Microsoft Teams meetings chat should be used. In addition, enabling the "Access links (cloud attachments) in messages" premium setting will automatically collect any meeting note (loop) or loop or word documents created by the Facilitator agent. Copilot Retention Reference Table Data Type Microsoft Purview Retention Policy Location/Scope Copilot prompts and responses Microsoft Copilot experiences Copilot Memories (personalized saved information Copilot "remembers") Not supported Copilot Pages (AI-generated, user-editable documents) SharePoint classic and communications sites (Static Scopes only) Facilitator interactions in a Team meeting Teams chats Facilitator meeting notes (loop) OneDrive Accounts Facilitator created word/loop documents OneDrive Accounts Facilitator generated and assigned tasks Exchange mailboxes (Tasks with end dates only)Why UK Enterprise Cybersecurity Is Failing in 2026 (And What Leaders Must Change)
Enterprise cybersecurity in large organisations has always been an asymmetric game. But with the rise of AI‑enabled cyber attacks, that imbalance has widened dramatically - particularly for UK and EMEA enterprises operating complex cloud, SaaS, and identity‑driven environments. Microsoft Threat Intelligence and Microsoft Defender Security Research have publicly reported a clear shift in how attackers operate: AI is now embedded across the entire attack lifecycle. Threat actors use AI to accelerate reconnaissance, generate highly targeted phishing at scale, automate infrastructure, and adapt tactics in real time - dramatically reducing the time required to move from initial access to business impact. In recent months, Microsoft has documented AI‑enabled phishing campaigns abusing legitimate authentication mechanisms, including OAuth and device‑code flows, to compromise enterprise accounts at scale. These attacks rely on automation, dynamic code generation, and highly personalised lures - not on exploiting traditional vulnerabilities or stealing passwords. The Reality Gap: Adaptive Attackers vs. Static Enterprise Defences Meanwhile, many UK enterprises still rely on legacy cybersecurity controls designed for a very different threat model - one rooted in a far more predictable world. This creates a dangerous "Resilience Gap." Here is why your current stack is failing- and the C-Suite strategy required to fix it. 1. The Failure of Traditional Antivirus in the AI Era Traditional antivirus (AV) relies on static signatures and hashes. It assumes malicious code remains identical across different targets. AI has rendered this assumption obsolete. Modern malware now uses automated mutation to generate unique code variants at execution time, and adapts behaviour based on its environment. Microsoft Threat Intelligence has observed threat actors using AI‑assisted tooling to rapidly rewrite payload components, ensuring that every deployment looks subtly different. In this model, there is no reliable signature to detect. By the time a pattern exists, the attacker has already moved on. Signature‑based detection is not just slow - it is structurally misaligned with AI‑driven attacks. The Risk: If your security relies on "recognising" a threat, you are already breached. By the time a signature exists, the attacker has evolved. The C-Suite Pivot: Shift investment from artifact detection to EDR/XDR (Extended Detection and Response). We must prioritise behavioural analytics and machine learning models that identify intent rather than file names. 2. Why Perimeter Firewalls Fail in a Cloud-First World Many UK enterprise still rely on firewalls enforcing static allow/deny rules based on IP addresses and ports. This model worked when applications were predictable and networks clearly segmented. Today, enterprise traffic is encrypted, cloud‑hosted, API‑driven, and deeply integrated with SaaS and identity services. AI‑assisted phishing campaigns abusing OAuth and device‑code flows demonstrate this clearly. From a network perspective, everything looks legitimate: HTTPS traffic to trusted identity providers. No suspicious port. No malicious domain. Yet the attacker successfully compromises identity. The Risk: Traditional firewalls are "blind" to identity-based breaches in cloud environments. The C-Suite Pivot: Move to Identity-First Security. Treat Identity as the new Control Plane, integrating signals like user risk, device health, and geolocation into every access decision. 3. The Critical Weakness of Single-Factor Authentication Despite clear NCSC guidance, single-factor passwords remain a common vulnerability in legacy applications and VPNs. AI-driven credential abuse has changed the economics of these attacks. Threat actors now deploy adaptive phishing campaigns that evolve in real-time. Microsoft has observed attackers using AI to hyper-target high-value UK identities- specifically CEOs, Finance Directors, and Procurement leads. The Risk: Static passwords are now the primary weak link in UK supply chain security. The C-Suite Pivot: Mandate Phishing‑resistant MFA (Passkeys or hardware security keys). Implement Conditional Access policies that evaluate risk dynamically at the moment of access, not just at login. Legacy Security vs. AI‑Era Reality 4. The Inherent Risk of VPN-Centric Security VPNs were built on a flawed assumption: that anyone "inside" the network is trustworthy. In 2026, this logic is a liability. AI-assisted attackers now use automation to map internal networks and identify escalation paths the moment they gain VPN access. Furthermore, Microsoft has tracked nation-state actors using AI to create synthetic employee identities- complete with fake resumes and deepfake communication. In these scenarios, VPN access isn't "hacked"; it is legally granted to a fraudster. The Risk: A compromised VPN gives an attacker the "keys to the kingdom." The C-Suite Pivot: Transition to Zero Trust Architecture (ZTA). Access must be explicit, scoped to the specific application, and continuously re‑evaluated using behavioural signals. 5. Data: The High-Velocity Target Sensitive data sitting unencrypted in legacy databases or backups is a ticking time bomb. In the AI era, data discovery is no longer a slow, manual process for a hacker. Attackers now use AI to instantly analyse your directory structures, classify your files, and prioritise high-value data for theft. Unencrypted data significantly increases your "blast radius," turning a containable incident into a catastrophic board-level crisis. The Risk: Beyond the technical breach, unencrypted data leads to massive UK GDPR fines and irreparable brand damage. The C-Suite Pivot: Adopt Data-Centric Security. Implement encryption by default, classify data while adding sensitivity labels and start board-level discussions regarding post‑quantum cryptography (PQC) to future-proof your most sensitive assets. 6. The Failure of Static IDS Traditional Intrusion Detection Systems (IDS) rely on known indicators of compromise - assuming attackers reuse the same tools and techniques. AI‑driven attacks deliberately avoid that assumption. Threat actors are now using Large Language Models (LLMs) to weaponize newly disclosed vulnerabilities within hours. While your team waits for a "known pattern" to be updated in your system, the attacker is already using a custom, AI-generated exploit. The Risk: Your team is defending against yesterday's news while the attacker is moving at machine speed. The C-Suite Pivot: Invest in Adaptive Threat Detection. Move toward Graph‑based XDR platforms that correlate signals across email, endpoint, and cloud to automate investigation and response before the damage spreads. From Static Security to Continuous Security Closing Thought: Security Is a Journey, Not a Destination For UK enterprises, the shift toward adaptive cybersecurity is no longer optional - it is increasingly driven by regulatory expectation, board oversight, and accountability for operational resilience. Recent UK cyber resilience reforms and evolving regulatory frameworks signal a clear direction of travel: cybersecurity is now a board‑level responsibility, not a back‑office technical concern. Directors and executive leaders are expected to demonstrate effective governance, risk ownership, and preparedness for cyber disruption - particularly as AI reshapes the threat landscape. AI is not a future cybersecurity problem. It is a current force multiplier for attackers, exposing the limits of legacy enterprise security architectures faster than many organisations are willing to admit. The uncomfortable truth for boards in 2026 is that no enterprise is 100% secure. Intrusions are inevitable. Credentials will be compromised. Controls will be tested. The difference between a resilient enterprise and a vulnerable one is not the absence of incidents, but how risk is managed when they occur. In mature organisations, this means assuming breach and designing for containment: Access controls that limit blast radius Least privilege and conditional access restricting attackers to the smallest possible scope if an identity is compromised Data‑centric security using automated classification and encryption, ensuring that even when access is misused, sensitive data cannot be freely exfiltrated As a Senior Enterprise Cybersecurity Architect, I see this moment as a unique opportunity. AI adoption does not have to repeat the mistakes of earlier technology waves, where innovation moved fast and security followed years later. We now have a rare chance to embed security from day one - designing identity controls, data boundaries, automated monitoring, and governance before AI systems become business‑critical. When security is built in upfront, enterprises don’t just reduce risk - they gain the confidence to move faster and unlock AI’s value safely. Security is no longer a “department”. In the age of AI, it is a continuous business function - essential to preserving trust and maintaining operational continuity as attackers move at machine speed. References: Inside an AI‑enabled device code phishing campaign | Microsoft Security Blog AI as tradecraft: How threat actors operationalize AI | Microsoft Security Blog Detecting and analyzing prompt abuse in AI tools | Microsoft Security Blog Post-Quantum Cryptography | CSRC Microsoft Digital Defense Report 2025 | Microsoft https://www.ncsc.gov.uk/news/government-adopt-passkey-technology-digital-servicesAuthorization and Governance for AI Agents: Runtime Authorization Beyond Identity at Scale
Designing Authorization‑Aware AI Agents at Scale Enforcing Runtime RBAC + ABAC with Approval Injection (JIT) Microsoft Entra Agent Identity enables organizations to govern and manage AI agent identities in Copilot Studio, improving visibility and identity-level control. However, as enterprises deploy multiple autonomous AI agents, identity and OAuth permissions alone cannot answer a more critical question: “Should this action be executed now, by this agent, for this user, under the current business and regulatory context?” This post introduces a reusable Authorization Fabric—combining a Policy Enforcement Point (PEP) and Policy Decision Point (PDP)—implemented as a Microsoft Entra‑protected endpoint using Azure Functions/App Service authentication. Every AI agent (Copilot Studio or AI Foundry/Semantic Kernel) calls this fabric before tool execution, receiving a deterministic runtime decision: ALLOW / DENY / REQUIRE_APPROVAL / MASK Who this is for Anyone building AI agents (Copilot Studio, AI Foundry/Semantic Kernel) that call tools, workflows, or APIs Organizations scaling to multiple agents and needing consistent runtime controls Teams operating in regulated or security‑sensitive environments, where decisions must be deterministic and auditable Why a V2? Identity is necessary—runtime authorization is missing Entra Agent Identity (preview) integrates Copilot Studio agents with Microsoft Entra so that newly created agents automatically get an Entra agent identity, manageable in the Entra admin center, and identity activity is logged in Entra. That solves who the agent is and improves identity governance visibility. But multi-agent deployments introduce a new risk class: Autonomous execution sprawl — many agents, operating with delegated privileges, invoking the same backends independently. OAuth and API permissions answer “can the agent call this API?” They do not answer “should the agent execute this action under business policy, compliance constraints, data boundaries, and approval thresholds?” This is where a runtime authorization decision plane becomes essential. The pattern: Microsoft Entra‑Protected Authorization Fabric (PEP + PDP) Instead of embedding RBAC logic independently inside every agent, use a shared fabric: PEP (Policy Enforcement Point): Gatekeeper invoked before any tool/action PDP (Policy Decision Point): Evaluates RBAC + ABAC + approval policies Decision output: ALLOW / DENY / REQUIRE_APPROVAL / MASK This Authorization Fabric functions as a shared enterprise control plane, decoupling authorization logic from individual agents and enforcing policies consistently across all autonomous execution paths. Architecture (POC reference architecture) Use a single runtime decision plane that sits between agents and tools. What’s important here Every agent (Copilot Studio or AI Foundry/SK) calls the Authorization Fabric API first The fabric is a protected endpoint (Microsoft Entra‑protected endpoint required) Tools (Graph/ERP/CRM/custom APIs) are invoked only after an ALLOW decision (or approval) Trust boundaries enforced by this architecture Agents never call business tools directly without a prior authorization decision The Authorization Fabric validates caller identity via Microsoft Entra Authorization decisions are centralized, consistent, and auditable Approval workflows act as a runtime “break-glass” control for high-impact actions This ensures identity, intent, and execution are independently enforced, rather than implicitly trusted. Runtime flow (Decision → Approval → Execution) Here is the runtime sequence as a simple flow (you can keep your Mermaid diagram too). ```mermaid flowchart TD START(["START"]) --> S1["[1] User Request"] S1 --> S2["[2] Agent Extracts Intent\n(action, resource, attributes)"] S2 --> S3["[3] Call /authorize\n(Entra protected)"] S3 --> S4 subgraph S4["[4] PDP Evaluation"] ABAC["ABAC: Tenant · Region · Data Sensitivity"] RBAC["RBAC: Entitlement Check"] Threshold["Approval Threshold"] ABAC --> RBAC --> Threshold end S4 --> Decision{"[5] Decision?"} Decision -->|"ALLOW"| Exec["Execute Tool / API"] Decision -->|"MASK"| Masked["Execute with Masked Data"] Decision -->|"DENY"| Block["Block Request"] Decision -->|"REQUIRE_APPROVAL"| Approve{"[6] Approval Flow"} Approve -->|"Approved"| Exec Approve -->|"Rejected"| Block Exec --> Audit["[7] Audit & Telemetry"] Masked --> Audit Block --> Audit Audit --> ENDNODE(["END"]) style START fill:#4A90D9,stroke:#333,color:#fff style ENDNODE fill:#4A90D9,stroke:#333,color:#fff style S1 fill:#5B5FC7,stroke:#333,color:#fff style S2 fill:#5B5FC7,stroke:#333,color:#fff style S3 fill:#E8A838,stroke:#333,color:#fff style S4 fill:#FFF3E0,stroke:#E8A838,stroke-width:2px style ABAC fill:#FCE4B2,stroke:#999 style RBAC fill:#FCE4B2,stroke:#999 style Threshold fill:#FCE4B2,stroke:#999 style Decision fill:#fff,stroke:#333 style Exec fill:#2ECC71,stroke:#333,color:#fff style Masked fill:#27AE60,stroke:#333,color:#fff style Block fill:#C0392B,stroke:#333,color:#fff style Approve fill:#F39C12,stroke:#333,color:#fff style Audit fill:#3498DB,stroke:#333,color:#fff ``` Design principle: No tool execution occurs until the Authorization Fabric returns ALLOW or REQUIRE_APPROVAL is satisfied via an approval workflow. Where Power Automate fits (important for readers) In most Copilot Studio implementations, Agents calls Power Automate (agent flows), is the practical integration layer that calls enterprise services and APIs. Copilot Studio supports “agent flows” as a way to extend agent capabilities with low-code workflows. For this pattern, Power Automate typically: acquires/uses the right identity context for the call (depending on your tenant setup), and calls the /authorize endpoint of the Authorization Fabric, returns the decision payload to the agent for branching. Copilot Studio also supports calling REST endpoints directly using the HTTP Request node, including passing headers such as Authorization: Bearer <token>. Protected endpoint only: Securing the Authorization Fabric with Microsoft Entra For this V2 pattern, the Authorization Fabric must be protected using Microsoft Entra‑protected endpoint on Azure Functions/App Service (built‑in auth). Microsoft Learn provides the configuration guidance for enabling Microsoft Entra as the authentication provider for Azure App Service / Azure Functions. Step 1 — Create the Authorization Fabric API (Azure Function) Expose an authorization endpoint: HTTP Step 2 — Enable Microsoft Entra‑protected endpoint on the Function App In Azure Portal: Function App → Authentication Add identity provider → Microsoft Choose Workforce configuration (enterprise tenant) Set Require authentication for all requests This ensures the Authorization Fabric is not callable without a valid Entra token. Step 3 — Optional hardening (recommended) Depending on enterprise posture, layer: IP restrictions / Private endpoints APIM in front of the Function for rate limiting, request normalization, centralized logging (For a POC, keep it minimal—add hardening incrementally.) Externalizing policy (so governance scales) To make this pattern reusable across multiple agents, policies should not be hardcoded inside each agent. Instead, store policy definitions in a central policy store such as Cosmos DB (or equivalent configuration store), and have the PDP load/evaluate policies at runtime. Why this matters: Policy changes apply across all agents instantly (no agent republish) Central governance + versioning + rollback becomes possible Audit and reporting become consistent across environments (For the POC, a single JSON document per policy pack in Cosmos DB is sufficient. For production, add versioning and staged rollout.) Store one PolicyPack JSON document per environment (dev/test/prod). Include version, effectiveFrom, priority for safe rollout/rollback. Minimal decision contract (standard request / response) To keep the fabric reusable across agents, standardize the request payload. Request payload (example) Decision response (deterministic) Example scenario (1 minute to understand) Scenario: A user asks a Finance agent to create a Purchase Order for 70,000. Even if the user has API permission and the agent can technically call the ERP API, runtime policy should return: REQUIRE_APPROVAL (threshold exceeded) trigger an approval workflow execute only after approval is granted This is the difference between API access and authorized business execution. Sample Policy Model (RBAC + ABAC + Approval) This POC policy model intentionally stays simple while demonstrating both coarse and fine-grained governance. 1) Coarse‑grained RBAC (roles → actions) FinanceAnalyst CreatePO up to 50,000 ViewVendor FinanceManager CreatePO up to 100,000 and/or approve higher spend 2) Fine‑grained ABAC (conditions at runtime) ABAC evaluates context such as region, classification, tenant boundary, and risk: 3) Approval injection (Agent‑level JIT execution) For higher-risk/high-impact actions, the fabric returns REQUIRE_APPROVAL rather than hard deny (when appropriate): How policies should be evaluated (deterministic order) To ensure predictable and auditable behavior, evaluate in a deterministic order: Tenant isolation & residency (ABAC hard deny first) Classification rules (deny or mask) RBAC entitlement validation Threshold/risk evaluation Approval injection (JIT step-up) This prevents approval workflows from bypassing foundational security boundaries such as tenant isolation or data sovereignty. Copilot Studio integration (enforcing runtime authorization) Copilot Studio can call external REST APIs using the HTTP Request node, including passing headers such as Authorization: Bearer <token> and binding response schema for branching logic. Copilot Studio also supports using flows with agents (“agent flows”) to extend capabilities and orchestrate actions. Option A (Recommended): Copilot Studio → Agent Flow (Power Automate) → Authorization Fabric Why: Flows are a practical place to handle token acquisition patterns, approval orchestration, and standardized logging. Topic flow: Extract user intent + parameters Call an agent flow that: calls /authorize returns decision payload Branch in the topic: If ALLOW → proceed to tool call If REQUIRE_APPROVAL → trigger approval flow; proceed only if approved If DENY → stop and explain policy reason Important: Tool execution must never be reachable through an alternate topic path that bypasses the authorization check. Option B: Direct HTTP Request node to Authorization Fabric Use the Send HTTP request node to call the authorization endpoint and branch using the response schema. This approach is clean, but token acquisition and secure secretless authentication are often simpler when handled via a managed integration layer (flow + connector). AI Foundry / Semantic Kernel integration (tool invocation gate) For Foundry/SK agents, the integration point is before tool execution. Semantic Kernel supports Azure AI agent patterns and tool integration, making it a natural place to enforce a pre-tool authorization check. Pseudo-pattern: Agent extracts intent + context Calls Authorization Fabric Enforces decision Executes tool only when allowed (or after approval) Telemetry & audit (what Security Architects will ask for) Even the best policy engine is incomplete without audit trails. At minimum, log: agentId, userUPN, action, resource decision + reason + policyIds approval outcome (if any) correlationId for downstream tool execution Why it matters: you now have a defensible answer to: “Why did an autonomous agent execute this action?” Security signal bonus: Denials, unusual approval rates, and repeated policy mismatches can also indicate prompt injection attempts, mis-scoped agents, or governance drift. What this enables (and why it scales) With a shared Authorization Fabric: Avoid duplicating authorization logic across agents Standardize decisions across Copilot Studio + Foundry agents Update governance once (policy change) and apply everywhere Make autonomy safer without blocking productivity Closing: Identity gets you who. Runtime authorization gets you whether/when/how. Copilot Studio can automatically create Entra agent identities (preview), improving identity governance and visibility for agents. But safe autonomy requires a runtime decision plane. Securing that plane as an Entra-protected endpoint is foundational for enterprise deployments. In enterprise environments, autonomous execution without runtime authorization is equivalent to privileged access without PIM—powerful, fast, and operationally risky.Microsoft Purview Data Quality Thresholds: More Control, More Trust
What Are Data Quality Thresholds? A data quality threshold defines the minimum acceptable score for a rule to pass. Instead of applying a single fixed standard across all data, organizations can now set expectations that align with business context and criticality. For example: An email column may require 99% completeness A product description column may only require 85% completeness Financial or regulatory data may require 100% accuracy With customizable thresholds, quality expectations become more meaningful and business-aligned. Why Does This Matter? Previously, using a single hardcoded threshold could lead to misleading quality scores. Critical data might appear “healthy” even when it didn’t meet business standards. With Data Quality Thresholds, you can: Define rule-level expectations Align quality scores with business risk Increase trust in DQ reporting Improve governance decision-making Data Asset-Level Quality Threshold Users can define data quality thresholds at the data asset level to measure how suitable a dataset is for specific business use cases. This allows organizations to quantify the overall health and fitness of a data asset before it is used in analytics, reporting, or data products. If the measured data quality score falls below the predefined threshold, the system can trigger notifications to the data asset owner or steward, prompting them to take corrective actions. It is important to note that not all data assets are equally critical. Therefore, thresholds should be context-driven and use-case specific. Example Scenario A marketing dataset used for campaign analysis may tolerate a lower quality threshold (e.g., 80%), since minor inconsistencies may not significantly impact insights. However, a financial reporting dataset used for regulatory filings may require a very high threshold (e.g., 98–100%), as even small errors can lead to compliance risks. Data Quality Rule-Level Threshold Thresholds can also be defined at the individual rule level, particularly for rules applied to specific columns. This provides more granular control and ensures that critical data elements are held to higher standards. Not all attributes have the same importance, so thresholds should reflect business criticality. Example Scenarios Email vs. Gender (Customer Contact Data) A completeness rule for a customer’s email address should have a higher threshold (e.g., 95–100%), since missing or invalid email addresses directly impact communication and engagement. In contrast, a gender attribute may have a lower threshold (e.g., 70–80%), as it is often less critical for most use cases. Billing Address vs. CRM Address A billing address is highly critical because it directly impacts: Invoice generation Tax calculations Timely delivery of invoices Therefore, the threshold for billing address quality should be very high (e.g., 98–100%). On the other hand, a CRM address used for general customer profiling may have a lower threshold, as occasional inaccuracies may not significantly affect business operations. The Impact By enabling flexible, context-aware scoring, Data Quality Thresholds help organizations move beyond generic quality checks and toward business-driven data quality management. Summary Data Quality Thresholds define the minimum acceptable score for data quality rules, allowing organizations to move beyond a one-size-fits-all approach and align quality expectations with business context and criticality. Instead of using fixed thresholds, organizations can set custom thresholds based on how important the data is. For example, financial data may require near-perfect accuracy, while less critical fields can tolerate lower thresholds. Thresholds can be applied at two levels: Data Asset Level: Measures the overall fitness of a dataset for a specific use case. Critical datasets (e.g., financial reporting) require higher thresholds than less critical ones (e.g., marketing analytics). Rule Level: Applies to individual columns or rules, ensuring that critical attributes (e.g., email, billing address) have stricter quality requirements than less important ones. This approach improves: Alignment with business risk and priorities Trust in data quality reporting Governance decision-making Focus on high-impact data issues Overall, data quality thresholds enable more meaningful, context-aware, and business-driven data quality management, helping organizations prioritize what matters most and build confidence in their data.
Optimizing OneDrive Retention Policies with Administrative Units and Adaptive Scopes
A special thank you note to Ashwini_Anand for contributing to the content of this blog. In today's digital landscape, efficient data retention management is a critical priority for organizations of all sizes. Organizations can optimize their OneDrive retention policies, ensuring efficient and compliant data management tailored to their unique user base and licensing arrangements. Scenario: Contoso Org encountered a distinct challenge - managing data retention for their diverse user base of 200,000 employees, which includes 80,000 users with F3 licenses and 120,000 users with E3 and E5 licenses. As per Microsoft licensing, F3 users are allocated only 2 GB of OneDrive storage, whereas E3 and E5 users are provided with a much larger allocation of 5 TB. This difference required creating separate retention policies for these users' groups. The challenge was further complicated by the fact that retention policies utilize the same storage for preserving deleted data. If a unified retention policy were applied to all users such as retaining data for 6 years before deletion - F3 users’ OneDrive storage could potentially fill up within a year or less (depending on usage patterns). This would leave F3 users unable to delete or save new files, severely disrupting productivity and data management. To address this, it is essential to create a separate retention policy for E3 and E5 users, ensuring that the policy applies only to these users and excludes F3 users. This blog will discuss the process of designing and implementing such a policy for the large user base based on separate licenses, ensuring efficient data management and uninterrupted productivity. Challenges with Retention Policy Configuration for large organizations 1. Adaptive Scope Adaptive scopes in Microsoft Purview allow you to dynamically target policies based on specific attributes or properties such as department, location, email address, custom Exchange attributes etc. Refer the link to get the list of supported attributes: Adaptive scopes | Microsoft Learn. Limitation: Although Adaptive scopes can filter by user properties, Contoso, being a large organization, had already utilized all 15 custom attributes for various purposes. Additionally, user attributes also couldn’t be used to segregate users based on licenses. This made it challenging to repurpose any attribute for our filter criteria to apply the retention policy to a specific set of users. Furthermore, refinable strings used in SharePoint do not work for OneDrive sites. 2. Static Scope Static scope refers to manually selected locations (e.g., specific users, mailboxes, or sites) where the policy is applied. The scope remains fixed and does not automatically adjust. Limitation: Static scope allows the inclusion or exclusion of mailboxes and sites but is limited to 100 sites and 1000 mailboxes, making it challenging to utilize for large organizations. Proposed Solution: Administrative Units with Adaptive Scope To address the above challenges, it required utilizing Administrative Units (Admin Units - is a container within an organization that can hold users, groups, or devices. It helps us to manage and organize users within an organization more efficiently, especially in large or complex environments) with Adaptive Scopes for creation of a retention policy targeting E3 and E5 licensed users. This approach allows organizations to selectively apply retention policies based on user licenses, enhancing both efficiency and governance. Prerequisites For Administrative unit - Microsoft Entra ID P1 license For Retention policy - Refer to the link: Microsoft 365 guidance for security & compliance - Service Descriptions | Microsoft Learn Configuration Steps Step 1: Create Administrative Unit: Navigate to Microsoft Entra Admin Center https://entra.microsoft.com/#home Click on ‘Identity’ and then click on ‘Show more’ Expand ‘Roles & admins’ Proceed to ‘Admin units’ -> Add. Figure 1: Create an Administrative unit and enter the name and description Define a name for the Administrative unit. Click on ‘Next: Assign roles’ No role assignment required, click on 'Next: Review + create’) Click on ‘Create’. To get more information about creating administrative unit, refer this link: Create or delete administrative units - Microsoft Entra ID | Microsoft Learn Step 2: Update Dynamic Membership: Select the Administrative Unit which is created in Step1. Navigate to ‘Properties’ Choose ‘Dynamic User’ for Membership type. Click on ‘Add a dynamic query’ for Dynamic user members. Click on ‘Edit' for Rule syntax In order to include E3 and E5 licensed users who are using OneDrive, you need to include SharePoint Online Service Plan 2 enabled users. Use the query below in the code snippet to define the dynamic membership. user.assignedPlans -any (assignedPlan.servicePlanId -eq "5dbe027f-2339-4123-9542-606e4d348a72" -and assignedPlan.capabilityStatus -eq "Enabled") 7. Click on 'Save' to update the Dynamic membership rules 8. Click on 'Save' to update the Administrative unit changes. 9. Open the Administrative Unit and click on the 'Users' tab to check if users have started to populate. Note: It may take some time to replicate all users, depending on the size of your organization. Please wait for minutes and then check again. Step 3: Create Adaptive Scope under Purview Portal: Access https://purview.microsoft.com Navigate to ‘Settings’ Expand ‘Roles & scopes’ and click on ‘Adaptive scopes’ Create a new adaptive scope, providing ‘Name’ and ‘Description’. Proceed to select the Administrative unit which was created earlier. (It takes time for the Admin/Administrative Unit to become visible. Please wait for some time if it does not appear immediately.) Click on ‘Add’ and ‘Next’ Select ‘Users’ and 'Next' Once the Admin unit is selected, we need to specify the criteria which allows to select users within the Admin unit (this is the second level of filtering available). However, in this case since we needed to select all users of the admin unit, hence the below criteria was used. Click 'Add attribute' and form the below query. Email addresses is not equal to $null Note: You can apply any other filter if you need to select a subset of users within the Admin Unit based on your business use case. Click on ‘Next’ Review and ‘Submit’ the adaptive scope. Step 4: Create Retention Policy using Adaptive Scope: Access to the portal https://purview.microsoft.com/datalifecyclemanagement/overview Navigate to ‘Policies’ and then go to ‘Retention Policies’. Create a ‘New Retention policy’, providing a ‘Name’ and ‘Description’. Click on "Next", there is no need to add Admin units here as its already defined in Adaptive scope. Figure 9: Select the 'Admin Units' as Full directory 6. Choose ‘Adaptive’ and click on ‘Next’. Click on ‘Add scopes’ and Select the previously created Adaptive scope. Under Location, select OneDrive. Figure 11: Select the Adaptive scope and location at this point. 8. Click on ‘Next’ to proceed and select the desired retention settings. 9. Click Next and Finish Outcome By implementing Admin Units with adaptive scopes, organizations can effectively overcome challenges associated with applying OneDrive retention policies for distinguished and large set of users. This approach facilitates the dynamic addition of required users, eliminating the need for custom attributes and manual user management. Users are dynamically added or removed from the policy based on license status, ensuring seamless compliance management. FAQ: Why is it important to differentiate retention policies based on user licensing tiers? It is important to differentiate retention policies based on user licensing tiers to ensure that each user group has policies tailored to their specific needs and constraints, avoiding issues such as storage limitations for users with lower-tier licenses like F3. How many Exchange custom attributes are typically available? There are typically 15 Exchange custom attributes available, which can limit scalability when dealing with a large user base. What challenge does Adaptive Scoping face when including a large number of OneDrive sites? Adaptive Scoping faces the challenge of including a large number of OneDrive sites due to limitations in the number of custom attributes allowed. While these custom attributes help in categorizing and managing OneDrive sites, the finite number of attributes available can restrict scalability and flexibility. Why are refinable strings a limitation for Adaptive Scoping in OneDrive? Refinable strings are a limitation for Adaptive Scoping in OneDrive because their usage is restricted to SharePoint only. What are the limitations of Static Scoping for OneDrive sites? Static Scoping for OneDrive sites is limited by the strict limit of including or excluding only 100 sites, making it usage limited for larger environments. Do we need any licenses to create an administrative unit with dynamic membership? Yes, a Microsoft Entra ID P1 license is required for all members of the group.Select the 'Adaptive' retention policy typeFigure 10: Select the 'Adaptive' retention policy type
Building Secure, Enterprise Ready AI Agents with Purview SDK and Agent Framework
At Microsoft Ignite, we announced the public preview of Purview integration with the Agent Framework SDK—making it easier to build AI agents that are secure, compliant, and enterprise‑ready from day one. AI agents are quickly moving from demos to production. They reason over enterprise data, collaborate with other agents, and take real actions. As that happens, one thing becomes non‑negotiable: Governance has to be built in. That’s where Purview SDK comes in. Agentic AI Changes the Security Model Traditional apps expose risks at the UI or API layer. AI agents are different. Agents can: Process sensitive enterprise data in prompts and responses Collaborate with other agents across workflows Act autonomously on behalf of users Without built‑in controls, even a well‑designed agent can create compliance gaps. Purview SDK brings Microsoft’s enterprise data security and compliance directly into the agent runtime, so governance travels with the agent—not after it. What You Get with Purview SDK + Agent Framework This integration delivers a few key things developers and enterprises care about most: Inline Data Protection Evaluate prompts and responses against Data Loss Prevention (DLP) policies in real time. Content can be allowed or blocked automatically. Built‑In Governance Send AI interactions to Purview for audit, eDiscovery, communication compliance, and lifecycle management—without custom plumbing. Enterprise‑Ready by Design Ship agents that meet enterprise security expectations from the start, not as a follow‑up project. All of this is done natively through Agent Framework middleware, so governance feels like part of the platform—not an add‑on. How Enforcement Works (Quickly) When an agent runs: Prompts and responses flow through the Agent Framework pipeline Purview SDK evaluates content against configured policies A decision is returned: allow, redact, or block Governance signals are logged for audit and compliance This same model works for: User‑to‑agent interactions Agent‑to‑agent communication Multi‑agent workflows Try It: Add Purview SDK in Minutes Here’s a minimal Python example using Agent Framework: That’s it! From that point on: Prompts and responses are evaluated against Purview policies setup within the enterprise tenant Sensitive data can be automatically blocked Interactions are logged for governance and audit Designed for Real Agent Systems Most production AI apps aren’t single‑agent systems. Purview SDK supports: Agent‑level enforcement for fine‑grained control Workflow‑level enforcement across orchestration steps Agent‑to‑agent governance to protect data as agents collaborate This makes it a natural fit for enterprise‑scale, multi‑agent architectures. Get Started Today You can start experimenting right away: Try the Purview SDK with Agent Framework Follow the Microsoft Learn docs to configure Purview SDK with Agent Framework. Explore the GitHub samples See examples of policy‑enforced agents in Python and .NET. Secure AI, Without Slowing It Down AI agents are quickly becoming production systems—not experiments. By integrating Purview SDK directly into the Agent Framework, Microsoft is making governance a default capability, not a deployment blocker. Build intelligent agents. Protect sensitive data. Scale with confidence.Monitor logical disk space through Intune
Hi All, We have a requirement to monitor low disk space, particularly on devices with less than 1GB of available space. We were considering creating a custom compliance policy, but this would lead to blocking access to company resources as soon as the device becomes non-compliant. Therefore, we were wondering if there are any other automated methods we could use to monitor the logical disk space (primarily the C drive) using Intune or Microsoft Graph. Thanks in advance, DilanLearn more about Microsoft Security Communities.
In the last five years, Microsoft has increased the emphasis on community programs – specifically within the security, compliance, and management space. These communities fall into two categories: Public and Private (or NDA only). In this blog, we will share a breakdown of each community and how to join.Secure and govern AI apps and agents with Microsoft Purview
The Microsoft Purview family is here to help you secure and govern data across third party IaaS and Saas, multi-platform data environment, while helping you meet compliance requirements you may be subject to. Purview brings simplicity with a comprehensive set of solutions built on a platform of shared capabilities, that helps keep your most important asset, data, safe. With the introduction of AI technology, Purview also expanded its data coverage to include discovering, protecting, and governing the interactions of AI apps and agents, such as Microsoft Copilots like Microsoft 365 Copilot and Security Copilot, Enterprise built AI apps like Chat GPT enterprise, and other consumer AI apps like DeepSeek, accessed through the browser. To help you view, investigate interactions with all those AI apps, and to create and manage policies to secure and govern them in one centralized place, we have launched Purview Data Security Posture Management (DSPM) for AI. You can learn more about DSPM for AI here with short video walkthroughs: Learn how Microsoft Purview Data Security Posture Management (DSPM) for AI provides data security and compliance protections for Copilots and other generative AI apps | Microsoft Learn Purview capabilities for AI apps and agents To understand our current set of capabilities within Purview to discover, protect, and govern various AI apps and agents, please refer to our Learn doc here: Microsoft Purview data security and compliance protections for Microsoft 365 Copilot and other generative AI apps | Microsoft Learn Here is a quick reference guide for the capabilities available today: Note that currently, DLP for Copilot and adhering to sensitivity label are currently designed to protect content in Microsoft 365. Thus, Security Copilot and Copilot in Fabric, along with Copilot studio custom agents that do not use Microsoft 365 as a content source, do not have these features available. Please see list of AI sites supported by Microsoft Purview DSPM for AI here Conclusion Microsoft Purview can help you discover, protect, and govern the prompts and responses from AI applications in Microsoft Copilot experiences, Enterprise AI apps, and other AI apps through its data security and data compliance solutions, while allowing you to view, investigate, and manage interactions in one centralized place in DSPM for AI. Follow up reading Check out the deployment guides for DSPM for AI How to deploy DSPM for AI - https://aka.ms/DSPMforAI/deploy How to use DSPM for AI data risk assessment to address oversharing - https://aka.ms/dspmforai/oversharing Address oversharing concerns with Microsoft 365 blueprint - aka.ms/Copilot/Oversharing Explore the Purview SDK Microsoft Purview SDK Public Preview | Microsoft Community Hub (blog) Microsoft Purview documentation - purview-sdk | Microsoft Learn Build secure and compliant AI applications with Microsoft Purview (video) References for DSPM for AI Microsoft Purview data security and compliance protections for Microsoft 365 Copilot and other generative AI apps | Microsoft Learn Considerations for deploying Microsoft Purview AI Hub and data security and compliance protections for Microsoft 365 Copilot and Microsoft Copilot | Microsoft Learn Block Users From Sharing Sensitive Information to Unmanaged AI Apps Via Edge on Managed Devices (preview) | Microsoft Learn as part of Scenario 7 of Create and deploy a data loss prevention policy | Microsoft Learn Commonly used properties in Copilot audit logs - Audit logs for Copilot and AI activities | Microsoft Learn Supported AI sites by Microsoft Purview for data security and compliance protections | Microsoft Learn Where Copilot usage data is stored and how you can audit it - Microsoft 365 Copilot data protection and auditing architecture | Microsoft Learn Downloadable whitepaper: Data Security for AI Adoption | Microsoft Explore the roadmap for DSPM for AI Public roadmap for DSPM for AI - Microsoft 365 Roadmap | Microsoft 365PMPur
