Microsoft critical vulnerabilities double: Exposure to escalation

Microsoft disclosed 1,273 vulnerabilities in 2025, a slight decrease from 1,360 the previous year, according to the 2026 Microsoft Vulnerabilities Report from BeyondTrust. At first glance, total vulnerability counts have remained relatively stable since 2020, which might seem like positive news.

However, that headline number is misleading. The real story lies in the surge of critical vulnerabilities, which doubled from 78 to 157 year-over-year, reversing a multi-year downward trend. This stability in total volume masks a dangerous escalation in potential impact, and that is exactly where organizations need to focus their attention.

The key insight is not how many flaws were disclosed, but where they are concentrated and what they enable attackers to do.

Where Risk Is Concentrating

Elevation of Privilege vulnerabilities now account for 40% of all CVEs, while Information Disclosure flaws jumped 73%. This shift signals that attackers are prioritizing stealth and reconnaissance over noisy, disruptive exploits. Privilege escalation is the tipping point where vulnerabilities become actual breaches. Threat actors no longer need loud malware campaigns when they can quietly escalate access and move laterally using legitimate credentials and Living Off the Land techniques.

This pattern matches real-world breach data, where initial access is often mundane, but the damage is amplified through excessive privilege, misconfigurations, and weak identity controls.

The most alarming concentration is in cloud and business platforms. While total vulnerabilities in Microsoft Azure and Dynamics 365 dipped slightly, critical flaws skyrocketed from 4 to 37 in a single year. Cloud platforms are no longer just infrastructure; they are the backbone of identity management, business automation, and enterprise control planes. A critical flaw in these environments can cripple entire workflows and collapse trust boundaries at machine speed. When a cloud vulnerability turns critical, the blast radius becomes the defining risk metric.

Consider CVE-2025-55241, a critical Entra ID flaw patched in July 2025. An attacker could forge tokens accepted across any tenant, leaving no trace in victim logs. A single misconfigured identity in Azure can hand an attacker the keys to your entire tenant, and most organizations wouldn’t know until the damage was done.

On the endpoint and server side, the results are mixed but still troubling. Microsoft Windows vulnerabilities declined overall, yet critical counts remained stubbornly high. Microsoft Windows Server vulnerabilities increased to 780, with 50 classified as critical. Servers remain high-value targets because they often run with elevated privileges, host shared services, and underpin critical business infrastructure. Threat actors know that compromising a server provides faster, deeper access than compromising a desktop alone. CISOs frequently ask, “We patched everything critical, so why are we still getting breached?” This data provides the answer.

Perhaps the most dramatic shift is in productivity software. Microsoft Office vulnerabilities surged 234% year-over-year, rising from 47 to 157, with critical flaws jumping from 3 to 31,a tenfold increase. Office remains one of the most abused attack surfaces because it sits at the intersection of human behavior, daily operations, and business continuity. Macros, document sharing, preview panes, HTML rendering, new AI capabilities, and add-ins create a unique landscape for exploitation. When Office vulnerabilities spike, users remain the most reliable entry point via social engineering.

These category trends reinforce a clear pattern: Elevation of Privilege and Information Disclosure are rising together. Attackers are prioritizing stealth and reconnaissance. When they know your environment better than your own team does, every subsequent incursion becomes easier.

What Organizations Should Do About It

The immediate defense priority is narrowing the blast radius before the next patch cycle. That means auditing standing admin rights, treating service accounts and AI agents with the same scrutiny as human identities, and disabling the Windows preview pane (seven CVEs in 2025 exploited it as an entry point).

Patch management alone is insufficient. Organizations must prioritize vulnerabilities that enable privilege escalation, identity abuse, and lateral movement first. This requires context, knowledge of exploits, and mappings to frameworks like MITRE ATT&CK, not just CVSS scores. It also requires rethinking trust assumptions across cloud, endpoint, server, and productivity layers.

The organizations that are ahead of this trend aren’t simply patching faster. They are rethinking what privilege means in a cloud-first environment. AI agents have quickly evolved from a future concern into a present reality, and most organizations lack the AI security posture management necessary for proper governance. Patches fail to fix excessive privilege or enforce least privilege for AI agents.

The ghost in this data isn’t the vulnerability count. It is everything those vulnerabilities unlock when identity controls aren’t there to stop them. For the 2026 landscape and beyond, the hard truth is clear: threat actors are not breaking down the front door with brute force exploits anymore. They are walking in, escalating quietly, and operating as trusted users,human and machine alike. If security programs don’t focus on privilege reduction, identity visibility, and continuous risk assessment, the numbers may look stable year-over-year, but the attack surface and business impact will continue to increase.