Microsoft closed out 2025 with patches for 56 safety flaws in varied merchandise throughout the Home windows platform, together with one vulnerability that has been actively exploited within the wild.
Of the 56 flaws, three are rated Important, and 53 are rated Vital in severity. Two different defects are listed as publicly identified on the time of the discharge. These embrace 29 privilege escalation, 18 distant code execution, 4 info disclosure, three denial-of-service, and two spoofing vulnerabilities.
In whole, Microsoft has addressed a complete of 1,275 CVEs in 2025, in line with information compiled by Fortra. Tenable’s Satnam Narang mentioned 2025 additionally marks the second consecutive 12 months the place the Home windows maker has patched over 1,000 CVEs. It is the third time it has performed so since Patch Tuesday’s inception.
The replace is along with 17 shortcomings the tech big patched in its Chromium-based Edge browser because the launch of the November 2025 Patch Tuesday replace. This additionally consists of a spoofing vulnerability in Edge for iOS (CVE-2025-62223, CVSS rating: 4.3).
The vulnerability that has come underneath energetic exploitation is CVE-2025-62221 (CVSS rating: 7.8), a use-after-free in Home windows Cloud Recordsdata Mini Filter Driver that would enable a licensed attacker to raise privileges regionally and acquire SYSTEM permissions.
“File system filter drivers, aka minifilters, connect to the system software program stack, and intercept requests focused at a file system, and lengthen or change the performance offered by the unique goal,” Adam Barnett, lead software program engineer at Rapid7, mentioned in a press release. “Typical use instances embrace information encryption, automated backup, on-the-fly compression, and cloud storage.”
“The Cloud Recordsdata minifilter is utilized by OneDrive, Google Drive, iCloud, and others, though as a core Home windows part, it might nonetheless be current on a system the place none of these apps had been put in.”
It is at the moment not identified how the vulnerability is being abused within the wild and in what context, however profitable exploitation requires an attacker to acquire entry to a vulnerable system by another means. Microsoft Menace Intelligence Middle (MSTIC) and Microsoft Safety Response Middle (MSRC) have been credited with discovering and reporting the flaw.
In response to Mike Walters, president and co-founder of Action1, a menace actor might acquire low-privileged entry by strategies like phishing, internet browser exploits, or one other identified distant code execution flaw, after which chain it with CVE-2025-62221 to grab management of the host.
Armed with this entry, the attacker might deploy kernel parts or abuse signed drivers to evade defenses and preserve persistence, and will be weaponized to realize a domain-wide compromise when coupled with credential theft situations.
The exploitation of CVE-2025-62221 has prompted the U.S. Cybersecurity and Infrastructure Safety Company (CISA) so as to add it to the Identified Exploited Vulnerabilities (KEV) catalog, mandating Federal Civilian Government Department (FCEB) businesses to use the patch by December 30, 2025.
The remaining two zero-days are listed beneath –
CVE-2025-54100 (CVSS rating: 7.8) – A command injection vulnerability in Home windows PowerShell that permits an unauthorized attacker to execute code regionally
CVE-2025-64671 (CVSS rating: 8.4) – A command injection vulnerability in GitHub Copilot for JetBrains that permits an unauthorized attacker to execute code regionally
“It is a command injection flaw in how Home windows PowerShell processes internet content material,” Action1’s Alex Vovk mentioned about CVE-2025-54100. “It lets an unauthenticated attacker execute arbitrary code within the safety context of a person who runs a crafted PowerShell command, akin to Invoke-WebRequest.”
“The menace turns into important when this vulnerability is mixed with widespread assault patterns. For instance, an attacker can use social engineering to steer a person or admin to run a PowerShell snippet utilizing Invoke-WebRequest, permitting a distant server to return crafted content material that triggers the parsing flaw and results in code execution and implant deployment.”
It is value noting that CVE-2025-64671 comes within the wake of a broader set of safety vulnerabilities collectively named IDEsaster that was just lately disclosed by safety researcher Ari Marzouk. The problems come up on account of including agentic capabilities to an built-in improvement atmosphere (IDE), exposing new safety dangers within the course of.
These assaults leverage immediate injections in opposition to the synthetic intelligence (AI) brokers embedded into IDEs and mix them with the bottom IDE layer to end in info disclosure or command execution.
“This makes use of an ‘previous’ assault chain of utilizing a susceptible instrument, so not precisely a part of the IDEsaster novel assault chain,” Marzouk, who’s credited with discovering and reporting the flaw, advised The Hacker Information. “Particularly, a susceptible ‘execute command’ instrument the place you may bypass the user-configured enable listing.”
Marzouk additionally mentioned a number of IDEs had been discovered susceptible to the identical assault, together with Kiro.dev, Cursor (CVE-2025-54131), JetBrains Junie (CVE-2025-59458), Gemini CLI, Windsurf, and Roo Code (CVE-2025-54377, CVE-2025-57771, and CVE-2025-65946). Moreover, GitHub Copilot for Visible Studio Code has been discovered to be vulnerable to the vulnerability, though, on this case, Microsoft assigned it a “Medium” severity ranking with no CVE.
“The vulnerability states that it is potential to realize code execution on affected hosts by tricking the LLM into operating instructions that bypass the guardrails and appending directions within the person’s ‘auto-approve’ settings,” Kev Breen, senior director of cyber menace analysis at Immersive, mentioned.
“This may be achieved by ‘Cross Immediate Injection,’ which is the place the immediate is modified not by the person however by the LLM brokers as they craft their very own prompts primarily based on the content material of information or information retrieved from a Mannequin Context Protocol (MCP) server that has risen in reputation with agent-based LLMs.”
Software program Patches from Different Distributors
Along with Microsoft, safety updates have additionally been launched by different distributors over the previous a number of weeks to rectify a number of vulnerabilities, together with —
Adobe
Amazon Internet Providers
AMD
Arm
ASUS
Atlassian
Bosch
Broadcom (together with VMware)
Canon
Cisco
Citrix
CODESYS
Dell
Devolutions
Django
Drupal
F5
Fortinet
Fortra
GitLab
Google Android and Pixel
Google Chrome
Google Cloud
Google Pixel Watch
Hitachi Power
HP
HP Enterprise (together with Aruba Networking and Juniper Networks)
IBM
Creativeness Applied sciences
Intel
Ivanti
Lenovo
Linux distributions AlmaLinux, Alpine Linux, Amazon Linux, Arch Linux, Debian, Gentoo, Oracle Linux, Mageia, Crimson Hat, Rocky Linux, SUSE, and Ubuntu
MediaTek
Mitsubishi Electrical
MongoDB
Moxa
Mozilla Firefox and Firefox ESR
NVIDIA
OPPO
Progress Software program
Qualcomm
React
Rockwell Automation
Samsung
SAP
Schneider Electrical
Siemens
SolarWinds
Splunk
Synology
TP-Hyperlink
WatchGuard
Zoom, and
Zyxel
