Netwrix has quietly rolled out a significant update to its 1Secure SaaS platform, embedding AI governance tools that target a sleeping giant in enterprise security: overly permissive access rights feeding Microsoft Copilot. The expansion, announced in a March 2025 product briefing, adds real-time monitoring of Copilot interactions, detection of sensitive data exposure through AI prompts, and a permissions risk engine that evaluates exactly what data a user’s access – not just the AI model – can surface. This move reframes the conversation around AI security from model-level guardrails to the foundational issue of identity and access hygiene.

For the roughly 70% of enterprises running hybrid Microsoft environments, the stakes are immediate. Copilot for Microsoft 365, embedded in Word, Excel, Teams, and Outlook, can instantly retrieve and synthesize information from SharePoint, OneDrive, and even on-premises file shares if connected through Azure AD. A single over-provisioned user account becomes a funnel through which sensitive data – salary sheets, merger plans, customer PII – flows into AI-generated summaries with no additional authorization checks. Netwrix’s new module attempts to plug that gap before breaches occur.

What Netwrix 1Secure Now Does

1Secure started as a cloud-delivered audit and compliance tool for hybrid Microsoft ecosystems, covering Active Directory, Azure AD, Exchange, SharePoint, and file servers. With this release, it extends into AI workload governance. The platform now ingests activity logs from Copilot for Microsoft 365 via the Microsoft Graph API and enriches them with identity context from on-prem AD and Entra ID (formerly Azure AD). Specifically, it:

  • Tracks every Copilot query and response where sensitive information appears, categorized by data type (financial, PII, IP) and source location.
  • Calculates a real-time permissions risk score for each user and group, weighting factors like excessive access rights, dormant accounts with seniority, or nested group memberships that grant unintended data reach.
  • Surfaces a “Copilot Exposure Report” that shows, for any given user, exactly which documents, emails, and conversations Copilot would index based on current permissions.
  • Automatically flags anomalous AI behavior, such as a salesperson suddenly querying HR records via Copilot, triggering an alert for manual review.

These additions build on 1Secure’s existing machine-learning-driven threat detection, which already identified unusual access patterns in Active Directory. By linking that capability to Copilot analytics, the platform creates a closed-loop system: permissions risks identified on the identity side directly translate into potential AI data leaks on the productivity side.

Why Permissions Risk Comes First

The product’s tagline – “permissions risk before Copilot model risk” – is a deliberate departure from the industry’s obsession with prompt injection and model safety. Netwrix’s own research, cited in the briefing, found that 63% of organizations have over 5% of sensitive files exposed to “everyone” or overly broad groups in SharePoint Online, and that misconfigured access control lists (ACLs) account for the majority of data breaches, not vulnerabilities in AI models themselves.

“We saw customers rolling out Copilot with great enthusiasm, only to realize their file permissions were a disaster. The AI just reflected what was already broken,” said Michael Tweddle, Netwrix’s VP of Product, during the announcement. “You can’t secure Copilot if you haven’t secured the data it touches. Our approach is to shine a light on those permission gaps first, then monitor the AI interactions for anomalies.”

This philosophy resonates with CISOs who often inherit decades of accumulated group memberships and open shares. For example, a marketing team member might retain access to a confidential M&A folder from a past project because nobody revoked it. With Copilot, a simple prompt like “summarize recent strategy documents” could pull that data without the user ever explicitly navigating to the folder. Netwrix’s Copilot Exposure Report gives security teams a preemptive view of what any user could potentially surface, allowing them to lock down access before turning on AI features.

Real-World Impact: A Hybrid Scenario

Consider a financial services firm with 5,000 users, Active Directory on-premises synchronized to Entra ID, and a mix of SharePoint Online and on-prem file servers. When Copilot is enabled, it indexes content across all locations the user can reach. A portfolio manager with access to internal trading strategies on an on-prem share might also have legacy read privileges to a SharePoint site containing unredacted client contracts. Through Copilot, that manager could unknowingly – or maliciously – query “show me contract terms for our top 10 clients” and receive a synthesis of sensitive data.

1Secure maps these hybrid data stores and cross-references them with each user’s effective permissions, displaying a visual graph of exposed assets. Alerts trigger when Copilot returns data from high-risk sources (HR databases, legal shares, finance folders) to a user whose role typically wouldn’t access them. The platform can also integrate with ServiceNow or Jira to automate remediation workflows, such as triggering an access review or temporarily disabling a user’s Copilot functionality.

Technical Underpinnings and Deployment

Netwrix 1Secure is a multi-tenant SaaS solution that requires a lightweight connector on premises to collect Active Directory and file server audit data. For Copilot, the integration uses Microsoft Graph API to pull user activity logs, including prompts, responses, and file references. The connection must be granted Organization.Read.All and AuditLog.Read.All API permissions in Entra ID, with audit logging enabled tenant-wide. Installation takes under an hour for most environments, according to Netwrix, and the AI governance features are available as an add-on module to existing 1Secure subscriptions.

The permissions risk engine uses a proprietary algorithm that factors in:

  • Static access assignments: direct ACLs, group memberships, and SharePoint permissions.
  • Dynamic access through conditional access policies that might grant temporary access.
  • Data sensitivity labels from Microsoft Purview (if in use), tagging files as Confidential, Highly Confidential, etc.
  • User behavior analytics to identify deviations from normal access patterns.

This multilayered approach means that even if a user’s permissions are technically correct but unusual for their job function, the system will raise a low-severity alert that can be escalated if combined with Copilot activity.

Market Context and Competition

The timing is strategic. Microsoft’s own Purview Compliance Manager and Defender for Cloud Apps offer some Copilot governance, such as data loss prevention (DLP) for prompts and responses. However, those tools are natively cloud-focused and often require E5 licensing, leaving hybrid shops with gaps. Third-party players like Varonis and SailPoint have also begun integrating AI governance features, but Netwrix’s strength lies in its deep Active Directory heritage and its ability to span on-prem and cloud in a single pane of glass.

Moreover, regulatory pressure is mounting. The SEC’s cybersecurity disclosure rules, EU AI Act, and industry-specific mandates like HIPAA and PCI DSS now explicitly require oversight of AI systems handling sensitive data. Netwrix’s auditing trails – which time-stamp every Copilot interaction and link it to a user’s identity and permissions – can be directly exported for compliance reports.

Early Adopter Feedback

Though the feature set is new, several beta customers shared experiences in the briefing. A healthcare provider with 12,000 users discovered that over 200 employees had inherited access to a shared drive containing patient records through a “Domain Users” group that had been granted read permissions years ago. The Copilot Exposure Report flagged this immediately, prompting a cleanup that reduced the organization’s permissions risk score by 40% within two weeks.

Another enterprise, a manufacturing company, used the anomalous behavior detection to catch a terminated contractor whose account hadn’t been fully deprovisioned; that account was used to query financial forecasts via Copilot in Excel, triggering an alert and an automated account disablement. “It was a wake-up call,” said the CISO. “We had focused so much on protecting the AI model that we forgot the basics of identity hygiene.”

The Road Ahead for AI Governance

Netwrix plans to expand the module with integration into Microsoft Copilot for Security and third-party AI tools like Salesforce’s Einstein GPT later in 2025. The next phase will introduce an AI-driven recommendation engine that automatically suggests permission adjustments based on actual Copilot usage patterns, effectively moving from detection to prevention.

In a landscape where enterprises scramble to adopt generative AI, the message is clear: the most dangerous security hole isn’t the AI model getting tricked; it’s the messy, over-trusted identity fabric underneath. Netwrix 1Secure’s shift toward AI governance acknowledges that the first step to securing Copilot is locking the doors it walks through.

For IT administrators and security teams, the immediate action item is to review permission hygiene before deploying Copilot broadly. Tools like Netwrix 1Secure provide visibility; the harder work is cultural – enforcing least privilege and regular access reviews. But with AI accelerating data accessibility, the cost of inaction could be a single search away from a headline-making breach.