Secure data as AI scales: New Microsoft Purview innovations at RSA 2026

Data estates are expanding fast—across Microsoft 365, endpoints, cloud platforms, and AI apps and agents. As this footprint grows, so do the opportunities for sensitive information to be overshared or exposed, especially when protection spans many tools and environments. The 2026 Data Security Index, Microsoft commissioned a survey of over 1,700 data security professionals concluded:

• 86% of leaders preferred integrated platforms over fragmented tools, citing better visibility, fewer alerts, and improved efficiency.
• 47% of organizations are implementing GenAI controls, up 8% from year over year.
• 82% of organizations have developed plans to use GenAI in their data security program.

To meet this moment, Microsoft Purview data security is evolving to cover a broader range of protections across apps, endpoints and cloud platforms. This integrated approach delivers consistent visibility, detection, and protection across the places data lives, moves, and is used – including AI interactions. In this blog, we’ll share how Purview helps organizations:

• Secure sensitive data in Microsoft 365 and endpoints
• Secure data across AI apps and agents
• Leverage AI-powered capabilities to identify and respond to data risks

This RSAC, we’re sharing new capabilities to help organizations discover sensitive data, apply necessary protections, and investigate data risks across their environments.

Secure data in Microsoft 365 and endpoints

A strong data security foundation begins with understanding where sensitive data lives and applying protections across the tools people use every day. Microsoft Purview delivers native data security across Microsoft 365, including SharePoint, Teams, Exchange, as well as endpoints like laptops and mobile devices, providing controls security teams need to protect data at scale.

What’s new at RSA 2026:

Customizable reporting for data security insights

Security teams today have to make sense of massive volumes of data security signals, yet it can be hard to quickly focus on the risks and metrics that matter most. Organizations need flexible ways to analyze signals, understand trends, and prioritize the issues that require attention. To address this, Microsoft Purview is introducing customizable reports, bringing flexible reporting to Purview for the first time. This new capability gives organizations greater control over how they view and understand their data security posture. Building on the out-of-the-box reports introduced at Microsoft Ignite, security teams can now tailor reporting by combining out-of-the-box reports and custom reports into views aligned to specific roles, priorities, or data security scenarios. Teams can drill into insights, analyze trends, and surface the signals that matter most for identifying risk, tuning policies, and strengthening their overall data security posture. Reports can also be exported for easy sharing, helping teams communicate insights with leadership, auditors, and stakeholders across the organization.

Figure 1: Security teams can create your own custom reports to track the metrics that matter most

Scaling labeling and data protection in SharePoint

As organizations prepare large data estates for AI, expanding labeling coverage helps ensure sensitive information is protected consistently across the organization. To help organizations scale their data security program, Purview now supports auto-labeling data-at-rest for SharePoint document libraries based on the library’s default sensitivity label. Building on existing behavior where new and updated files inherit the library’s default label, this capability now extends the same label inheritance to existing content stored in the library. When a default label is configured on a SharePoint library, files at rest within the library will be automatically labeled the same way—without relying on users to manually label content. This allows organizations to label and protect data at scale before it’s accessed by Microsoft 365 Copilot or shared broadly, reducing oversharing risks in collaboration and AI workflows.

Enhanced protection for Microsoft & macOS environments

• Policy tips enable organizations to balance enforcement of DLP protections with end user awareness & education, preventing contextless disruption to their productivity. Policy tips will now be supported in preview for a broader range of devices, including Outlook on macOS and mobile devices, in the coming weeks. For a full list of supported policy tips by location, visit Learn. Additionally, policy tips appearing in Edge for Business now support custom hyperlinks in preview. This augments the custom hyperlink support for Windows devices announced last November.
• To prevent end users from copying files with sensitive data to potentially risky network shares, Purview DLP can now restrict sensitive data transfer to network shares leveraging WebDAV protocol in preview on Windows and macOS devices.
• Last April, we introduced the ability for admins to leverage just-in-time protection to protect OneDrive shadow files that are not locally synced on an endpoint device. This capability will be available in preview for macOS devices in addition to Windows in the coming weeks.
• Purview Insider Risk Management (IRM) now supports Microsoft Fabric Lakehouse indicators, along with quick policies for data theft detection to detect risky activities across non-Microsoft 365 apps, including Microsoft Fabric from departing users or accounts deleted from Microsoft Entra ID. These indicators enable admins to create and customize insider risk policies that detect a broader set of risky activities across Fabric workloads. These capabilities are now generally available.

Improved Admin experience

• Last February, we introduced collection policies – a configuration option in the Purview portal to reduce activity log and detection noise across your tenant from classifiers or user activities that aren’t relevant to your organization. This capability, already supported for endpoint devices, enterprise AI, and browser & network traffic, can be leveraged to scope relevant data and activity detections in SharePoint and OneDrive in preview, in the coming weeks. 
• For confident deployment of DLP policies to organizational devices, the Devices dashboard in Purview DLP now confirms policies have been properly synced to devices within 30 minutes of policy application. Additionally, the dashboard now provides visibility into the user profile logged into each device. These improvements to the Devices dashboard are generally available.
• Last year, we made troubleshooting of Purview DLP for Windows devices more efficient for admins with the introduction of always-on diagnostics. By providing admins direct access to diagnostics logs from user devices directly in the Purview portal, admins can work with our support team to resolve technical issues while minimizing disruption to information workers. Always-on diagnostics is now generally available for macOS devices in addition to Windows.
• Insider Risk Management (IRM) supports the ability to preview content in Activity Explorer, facilitating faster triage by allowing investigators to validate risk the moment suspicious activity appears without creating a case or waiting for content to be downloaded. This capability will be available in public preview soon.
• Analysts now have the ability to create and scale IRM to up to 2,000 active cases without content download (can be enabled later), supporting faster responses when potential risk is identified and facilitating content download prioritization for the subset of scenarios that need deeper investigation. This capability is now in public preview.

Secure data across AI apps and agents

Microsoft Purview extends data security to AI interactions, applying protections that reduce risk, prevent exposure, and enable secure AI adoption.

What’s new at RSA 2026:

Protect sensitive data in AI interactions

• The new DLP for Microsoft 365 Copilot web search control, in public preview, enables you to selectively block Copilot prompts containing sensitive information types (SITs) from being sent to external web search, while still allowing Copilot to respond using permitted Microsoft 365 enterprise data when available. In contrast to the DLP control that blocks Copilot grounding and responses entirely when sensitive data is detected in the prompt, this enhancement targets only the web‑grounding path, enabling organizations to prevent data leaks via web queries without sacrificing productivity through Copilot. For example, if a user asks Copilot to analyze a scenario that includes a customer’s credit card number, the prompt will not be sent to Bing, but Copilot can still generate an answer based on internal SharePoint site if allowed.
• The introduction of DLP for Copilot Studio[1], in public preview, brings inline, real-time DLP controls to Copilot Studio agents by detecting sensitive information types directly in prompts sent to custom agents and blocking those prompts before the agent is invoked. Previously, protections were limited to honoring sensitivity labels in agent interactions and blocking interactions based on the sensitivity label of a file being used on the prompt. Now, sensitive data in the prompt itself can also be blocked during input, before it can be processed or propagated. This reduces the risk of sensitive data leaks and enables safe scaling of custom agents by preventing accidental or improper use of sensitive or regulated data in prompts—such as pasting real customer IDs into an agent prompt.

Figure 2: Prompt with sensitive information blocked from being processed by Copilot Studio

Learn more about how Purview helps secure data in AI workflows in this blog.

Extend protection to unmanaged and third party AI usage

• Last February, we announced our ecosystem of secure access service edge (SASE) partners to extend Purview data security capabilities to the network. We are pleased to announce that Purview now integrates with Palo Alto Networks Prisma SASE to detect and block sensitive data in transit over the network, such as sensitive data that may be sent in prompts & responses with unmanaged AI apps and agents. This integration will be available in preview in the coming weeks.
• In addition to integrating with Palo Alto Networks Prisma SASE for the network layer, Purview DLP is also introducing a new endpoint DLP browser extension to prevent uploads and copy+paste of sensitive data to Palo Alto Networks Prisma Browser. This safeguards against the sharing of organizational data to potentially risky websites, including unmanaged consumer AI sites, as defined by sensitive service domains. The endpoint DLP browser extension, augmenting existing coverage across Edge, Chrome, and Firefox, is now available in preview.
• Beyond the breadth of Purview data security controls built into Edge for Business, Purview now integrates directly into non-Edge enterprise browsers, starting with the Island Enterprise Browser and Island Extension. By integrating Island and Purview, Island can enforce data security controls directly in the browser as users type, paste, upload, or share information. These controls leverage the same Purview classifiers, sensitivity labels, and DLP policies that many organizations rely on across the rest of their digital estate, including Microsoft 365, to detect and block sensitive data in web traffic, even without onboarding the device itself to Purview DLP. This enables consistent protection for data shared with AI and web applications across the diverse range of browsers leveraged by today’s enterprises. This integration is now available in preview. Learn more here.
• The Microsoft Security Store is now accessible directly in the Purview portal. Through this embedded experience, you can now discover & deploy Purview integrations from Microsoft Entra Global Secure Access and partners such as Netskope, iBoss, Palo Alto Networks, and Island to extend Purview data security to networks and secure browsers. Purchases continue seamlessly through the Secure Store website for a consistent and secure transaction. Learn more.

Leverage AI-powered capabilities to identify and respond to data risks

Visibility into data risk is integral, but moving to action is key. Security teams are flooded with signals, alerts and complex investigations. Fragmented tools and triage processes slow response and make it harder to focus on what matters most. Microsoft Purview uses AI to connect signals across data, user activity and endpoints, helping teams prioritize risk, accelerate investigations and response.

What’s new at RSA 2026:

Investigate & respond with connected signals

• Microsoft Purview Data Security Investigations (DSI) now lets customers analyze content associated with user activities and insights You can now begin a Data Security Investigation by building a query for Unified Audit Log activities or endpoint Data Loss Prevention (DLP) alerts, automatically pulling the files associated with suspicious actions—such as downloads, access, or transfers—into a single investigation for AI-powered deep content analysis. By defining criteria like time range, users, devices, activities, or keywords, investigators can quickly answer questions like “What files did this user interact with before they left the company?” without manually correlating logs and content. This evolution, which allows content searches based on activities helps teams move faster, reduce guesswork, and focus on real risk when responding to potential data exfiltration. This capability will be in public preview in the coming weeks.
• The Data Security Posture Agent now includes a credential scanning capability, available in preview, to help security teams proactively identify credentials such as passwords, API keys, certificates, tokens, etc. exposed secrets within a user-defined scope. Admins select the people, groups, or locations to review, and the agent analyzes context and usage to detect credentials stored where they should not be. Findings are surfaced with risk levels, confidence scores, and AI-generated reasoning. This helps teams systematically uncover credential exposure, reduce manual investigation effort, and strengthen data security posture through informed, human-validated decisions.

Figure 3: Output of a credential scan performed by the Data Security Posture Agent

• The Data Security Triage Agent now includes an advanced AI reasoning layer that performs coordinated, multi‑step analysis across insider risk activity, user, and data signals in Insider Risk Management. The agent breaks down complex alerts, reasons over behavioral patterns and contextual indicators, and synthesizes those signals into a clearer risk narrative for each investigation. This results in more accurate classification and prioritization of insider risk alerts, with richer context surfaced upfront. This way, analysts can spend less time correlating logs and reconstructing activity timelines, and more time validating true risk and taking appropriate action with confidence. Learn more about improvements to the Insider Risk Management investigation experience here.
• The Data Security Triage Agent is also receiving an upgrade that makes custom Sensitive Information Types (SITs) easier for both the agent and analysts to understand in Data Loss Prevention alerts. Custom SITs are often difficult for analysts to interpret quickly because the underlying definitions and patterns lack clear context at triage time. With this new capability, Purview uses AI to interpret custom SIT definitions, generate semantic descriptions of what the data represents, and surface that context directly within the agent. As a result, the Data Security Triage Agent can more accurately classify and prioritize alerts involving custom data, helping analysts understand why an alert matters and focus on real risk faster.

 The innovations announced at RSA reinforce Microsoft Purview’s role in delivering native, AI powered data security – helping customers protect sensitive data across Microsoft environments, extending those protection controls to AI apps and agents, and using AI to accelerate investigation and response.

Learn more: Explore Microsoft Purview data security capabilities, review documentation or start a free trial to see how Purview can help you secure your data, wherever it lives and travels.

[1] Copilot Studio agents published through Microsoft 365 Copilot.