Patch Tuesday’s 79 flaws expose identity and runtime risk

At a glance

What this is: This Patch Tuesday roundup highlights 79 vulnerabilities, including privilege escalation, denial of service, and a zero-click Excel issue tied to Copilot data disclosure, plus active exploitation of an abandoned WordPress plugin.

Why it matters: It matters because identity, privilege, and runtime trust boundaries determine whether a vulnerability becomes a contained defect or a path to data exposure, server compromise, or AI-assisted leakage.

By the numbers:

• This month’s Patch Tuesday release includes 79 new CVEs, including three classified critical.

👉 Read Expel's analysis of March 2026 Patch Tuesday and active WordPress exploitation

Context

Patch management is not just a vulnerability hygiene exercise. It is a governance test for whether organisations can reduce exposure fast enough across platforms, applications, and identity-mediated access paths before attackers turn known flaws into privilege, persistence, or data loss.

This article sits at the intersection of broader cybersecurity and identity governance because the most damaging outcomes here involve authorization failure, unauthorised execution, and data disclosure through software that already has trusted access. That includes privileged database access, application runtime stability, and AI-assisted information leakage through Copilot-enabled workflows.

Key questions

Q: What breaks when unsupported software stays on a live production system?

A: Unsupported software creates a permanent exposure window because the code remains reachable even after the maintainer has stopped fixing flaws. Attackers look for these orphaned components because they often retain execution privileges, privileged file paths, or authentication shortcuts. Once discovered, a single bug can become remote code execution, data theft, or malware delivery.

Q: Why do privilege escalation flaws in database platforms matter so much?

A: Privilege escalation in a database is not just a local bug. It can convert an ordinary authenticated session into sysadmin control, which means the attacker can read, modify, or pivot through business-critical data. That makes authorization validation one of the most important identity controls in the database stack.

Q: How do security teams know whether Patch Tuesday exposure is actually shrinking?

A: Look beyond patch counts and measure how long critical flaws remain reachable in production, how quickly high-risk systems are remediated, and whether privileged paths are monitored for abuse. Exposure is shrinking only when the organisation can remove or contain dangerous flaws before they are actively exploited.

Q: Who is accountable when a zero-click disclosure path exposes sensitive data?

A: Accountability usually sits across application ownership, identity governance, and data protection teams because the failure spans document processing, access entitlement, and information disclosure. Where an AI assistant can surface data from hidden paths, organisations must review who approved access, who monitored the workflow, and who owns the risk acceptance decision.

Technical breakdown

How SQL Server privilege escalation becomes sysadmin control

SQL Server elevation-of-privilege flaws usually sit at the authorization layer, where a request is accepted and processed as though it came from a higher-privileged context. In this case, a remote authenticated attacker can send a specially crafted command that alters the engine’s authorization decision and elevates to sysadmin. Once that happens, the problem is no longer just a vulnerability in code. It becomes a database trust failure, because sysadmin-level access can read, modify, and potentially chain into broader application compromise.

Practical implication: tighten database authentication, monitor privileged command patterns, and treat SQL Server admin rights as high-risk identity paths.

Why zero-click Excel and Copilot create a disclosure path

Zero-click disclosure issues are especially concerning when an application’s internal processing interacts with an AI agent or assistant that can surface hidden data. Here, improper neutralization of input in Excel can cause a Copilot agent to disclose sensitive information over the network without a user deliberately executing a malicious payload. That changes the threat model from a classic endpoint exploit to a trust-boundary failure between document content, local processing, and AI-assisted output generation.

Practical implication: review AI assistant data flows, restrict what Copilot can access by default, and validate document-based trust boundaries.

How abandoned WordPress plugins turn into persistent attack surfaces

An abandoned plugin remains part of the attack surface as long as it is reachable, even if the original developer no longer supports it. The Kaswara flaw works because the plugin accepts uploaded ZIP content without proper authentication or file validation, then unpacks it into a web-accessible directory. That gives an attacker a direct route from unauthenticated upload to remote code execution. Once code runs, the server can be used for cryptomining, redirect injection, or further foothold expansion.

Practical implication: remove unsupported plugins, scan upload paths for executable files, and treat legacy extensions as active exposure until removed.

Threat narrative

Attacker objective: The attacker’s objective is to turn a forgotten plugin into remote code execution that enables cryptomining or website takeover.

1. Entry occurs when an attacker finds an exposed WordPress site still running the abandoned Kaswara plugin and submits a malicious ZIP payload through the vulnerable upload path.
2. Escalation follows when the plugin unpacks a PHP script into a publicly accessible directory, allowing the attacker to execute server-side code without authentication.
3. Impact occurs when the attacker uses that foothold to deploy XMRig cryptomining software or take control of the site for broader abuse.

Breaches seen in the wild

• MITRE ATT&CK Enterprise Matrix — MITRE ATT&CK Enterprise — adversary tactics and techniques, threat detection, attack chain mapping, credential access, lateral movement, privilege escalation.
• JetBrains GitHub plugin token exposure — CVE-2024-37051 in JetBrains IntelliJ GitHub plugin exposed GitHub access tokens.

Read our 52 NHI Breaches Analysis report for a comprehensive view of breaches impacting Non-Human Identities including AI Agents.

NHI Mgmt Group analysis

Patch volume is a governance signal, not an operating plan. Seventy-nine CVEs in a single release tells teams that prioritisation, not raw patch throughput, is the control that matters. The challenge is separating exposures that threaten privilege boundaries from those that mainly affect availability. In practice, security leaders should rank fixes by the identity or execution paths they can open, not by headline count alone.

Zero-click disclosure through Copilot shows that AI assistants inherit application trust problems. When an assistant can surface information from document processing paths, the issue becomes AI data governance as much as application security. That is where identity and NHI governance intersect, because assistant workflows often rely on service credentials, delegated permissions, and network-accessible data sources. Practitioners need to treat AI-enabled disclosure as an access-control problem, not just a content-filtering problem.

Abandoned software creates a standing control gap that standard patching cannot close. The Kaswara case illustrates a specific failure mode: unsupported components persist long after their owners have stopped maintaining them, yet they retain execution privileges inside live environments. This is a lifecycle governance problem that belongs in asset, identity, and application control reviews. If a component can still execute code and no one can patch it, the organisation has already lost control of that trust boundary.

Named concept: patch-to-exposure latency. The longer a known flaw stays reachable in production, the more likely it is to become an identity, data, or runtime incident rather than a vulnerability item. That latency is especially damaging where privileged services, AI assistants, or public web paths are involved. The practical conclusion is to measure exposure duration in business-critical systems, not just count open tickets.

From our research:

• Ultimate Guide to NHIs reports that 71% of NHIs are not rotated within recommended time frames, increasing the risk of compromise over time.
• Our research also shows that only 5.7% of organisations have full visibility into their service accounts, which makes orphaned access and stale credentials harder to spot.
• For a deeper lifecycle lens, read Ultimate Guide to NHIs , Key Challenges and Risks for the visibility, rotation, and offboarding failures that let exposure linger.

What this signals

Patch management is increasingly a control-plane problem, not a ticket-queue problem. When privileged services, web runtime paths, and AI assistants all sit on top of the same exposure stack, teams need to see vulnerabilities as access-risk events rather than isolated software defects.

Patch-to-exposure latency: the operational gap between disclosure and actual risk reduction is now a measurable governance issue. If unsupported plugins, privileged database paths, or AI-enabled document flows remain reachable, the organisation has not reduced risk, it has only recorded it.

Teams should expect more mixed incidents where application flaws and identity misuse reinforce each other. That means tighter asset inventories, better runtime visibility, and stronger linkage between vulnerability management, identity governance, and data protection workflows.

For practitioners

• Prioritise flaws that can cross trust boundaries Rank this Patch Tuesday release by whether each issue can lead to privilege escalation, unauthorised execution, or data disclosure. Focus first on the SQL Server and Copilot-adjacent flaws because they affect privileged access and sensitive information paths.
• Review AI assistant data paths for disclosure risk Map what Copilot can read, process, and return from office documents and connected services. Reduce unnecessary access, limit sensitive content exposure, and test whether neutralization failures can surface internal data over the network.
• Remove unsupported plugins from live WordPress estates Delete abandoned extensions like Kaswara immediately, then inspect upload directories such as wp-content/uploads/kaswara/fonts_icon/ for executable files. If any PHP files appear there, treat the site as potentially compromised and escalate to full audit.
• Audit privileged database rights and command patterns Review who can issue high-risk SQL Server commands, then alert on unusual authorization requests and admin-context operations. The goal is to detect attempts to turn an authenticated session into sysadmin control before data is altered.

Key takeaways

• This Patch Tuesday release is less about the number of CVEs than about which flaws can break trust boundaries and elevate access.
• The abandoned WordPress plugin example shows how unsupported software turns a known vulnerability into a persistent execution path.
• Teams should prioritise privileged access, AI-assisted disclosure paths, and legacy components that remain reachable in production.

Key terms

• Patch-to-exposure latency: The time between a vulnerability being disclosed and an organisation actually reducing the risk in production. It matters because a published flaw is only theoretical until a reachable system still exposes the vulnerable path. Long latency often signals weak asset visibility, slow remediation, or poor ownership.
• Zero-click disclosure: A data exposure condition where information can leak without the victim deliberately opening a file or approving an action. In practice, the problem sits at the boundary between content processing, application trust, and data access, making it especially dangerous when assistants or automated workflows are involved.
• Unsupported software exposure: Risk created when a live system continues to run software that no longer receives patches or security fixes. The component remains an active attack surface until it is removed, isolated, or replaced, which means lifecycle management is a security control, not just an inventory task.
• Privilege Escalation: An attack technique where a compromised identity — often an NHI with initially limited permissions — exploits vulnerabilities or misconfigurations to gain elevated access rights, typically leading to broader compromise.

What's in the full article

Expel's full analysis covers the operational detail this post intentionally leaves for the source:

• Walkthroughs of the SQL Server, .NET, and Excel vulnerability mechanics in the March 2026 release
• The specific exploit path used in the abandoned WordPress plugin case, including upload handling and execution flow
• Indicators and file-system checks for finding suspicious PHP files in WordPress upload directories
• Context on why the Copilot-linked Excel issue changes the disclosure model for document workflows

👉 Expel's full post covers the SQL Server flaw, Copilot disclosure risk, and Kaswara exploit path in more operational detail.

Deepen your knowledge

The NHI Foundation Level course, the industry's only accredited NHI security programme, covers NHI governance, secrets management, and workload identity. It helps practitioners connect identity control to the broader security programme they are responsible for.