On a Thursday in September 2026, GitLab disclosed a flaw so fundamental in its severity that it carries the highest possible danger rating — a single, unauthenticated HTTP request is all that separates an attacker from every file on a vulnerable server. The vulnerability, residing in the Commits API, requires no credentials, no insider knowledge, and no sophistication to exploit, placing it among the most democratically dangerous flaws in recent memory. Within hours of disclosure, the internet was already alive with probes, reminding us that in the modern threat landscape, the distance between
GitLab's Critical CVSS 10 Flaw Exploited in the Wild Hours After Disclosure
One HTTP request, every file on the server
So this is a path traversal flaw—what does that actually mean for someone running GitLab?
It means an attacker can ask the server for a file, and the server gives it to them. Normally, applications are designed to only let you access files in a specific folder. This flaw breaks that boundary.
But how does that work technically? Is it just a malformed URL, or is there something more sophisticated happening?
From what the researchers documented, it's a malformed request to the Commits API. You're essentially telling the server to traverse up and out of its normal directory structure and grab whatever file you want.
And the CVSS 10 score—that's the maximum. What makes this different from other critical vulnerabilities?
CVSS 10 means it's unauthenticated, requires no user interaction, and has complete impact on confidentiality. You don't need to be logged in. You don't need to trick anyone. You just send the request.
The source material says exploitation was detected "within hours." Do we know if that's hours after disclosure, or hours after the vulnerability was first discovered?
The reporting indicates it's hours after public disclosure. Once GitLab announced it, attackers started scanning for vulnerable instances immediately.
What's the actual blast radius here? How many GitLab instances are exposed?
That's the thing—we don't have a number. The reporting says probes were detected "across the internet" and "internet-wide," but there's no count of how many instances are actually vulnerable or have been compromised.
What we do know is that any organization running an unpatched version is at risk. The vulnerability affects the Commits API, so if you're exposing GitLab to the internet, you're exposed.
What files can attackers actually read?
Anything on the server. Configuration files, private keys, environment variables, source code, database credentials—whatever is stored there.
And we should note: the source material doesn't specify which versions of GitLab are affected or whether there's a workaround short of patching. That's important information that's missing from the reporting.
So the only real defense is to patch immediately?
Patch immediately, and assume that unpatched systems may have already been compromised. You need to review logs, rotate credentials, and look for signs of data theft.
The Pulse
- A CVSS 10 path traversal flaw in GitLab's Commits API lets any unauthenticated attacker read arbitrary server files with a single HTTP request — no account, no credentials, no complexity required.
- Active exploitation began within hours of public disclosure, with security researchers documenting live scanning and file extraction attempts against real GitLab instances across the internet.
- The files at risk are not trivial — database credentials, private keys, environment variables, and sensitive source code all sit within reach of a successful exploit.
- GitLab issued an urgent call to patch immediately, warning that every unpatched instance is effectively an open door to total compromise.
- Security teams are advised to treat any unpatched system as already breached — rotating credentials, reviewing access logs, and monitoring for data exfiltration alongside emergency patching.
On a Thursday in September 2026, GitLab disclosed a flaw so fundamental in its severity that it carries the highest possible danger rating — a single, unauthenticated HTTP request is all that separates an attacker from every file on a vulnerable server. The vulnerability, residing in the Commits API, requires no credentials, no insider knowledge, and no sophistication to exploit, placing it among the most democratically dangerous flaws in recent memory. Within hours of disclosure, the internet was already alive with probes, reminding us that in the modern threat landscape, the distance between announcement and exploitation has collapsed to nearly nothing.
GitLab disclosed a maximum-severity vulnerability on Thursday that carries a CVSS score of 10 — the highest possible rating. Tracked as CVE-2026-85706, the flaw lives in the Commits API and requires no authentication whatsoever. A single, specially crafted HTTP request is enough to bypass file access controls entirely, granting an attacker read access to any file on the server.
The attack's simplicity is what makes it so alarming. An attacker needs nothing more than a target URL. No GitLab account, no credentials, no deep knowledge of the system. The files exposed could include database passwords, private keys, environment variables, and sensitive source code — the kind of data that, once accessed, can lead to total infrastructure compromise.
What transformed this from a serious disclosure into a crisis was the speed of the response from the other side. Security researchers, including those at watchTowr, documented active exploitation attempts within hours of the vulnerability becoming public. Attackers were already scanning the internet for vulnerable instances and attempting to extract files before most organizations had even read the advisory. Multiple outlets — BleepingComputer, The Hacker News, CyberScoop — independently confirmed the rapid, widespread attack activity.
For organizations running GitLab, the path forward is urgent and unambiguous: patch immediately, treat any unpatched system as potentially compromised, rotate exposed credentials, and review access logs for signs of exfiltration. The window between disclosure and active exploitation was measured in minutes. The only meaningful defense left is speed.
GitLab disclosed a maximum-severity vulnerability on Thursday that allows attackers to read any file on a server with nothing more than a single HTTP request. The flaw, tracked as CVE-2026-85706 and assigned a CVSS score of 10—the highest possible rating—lives in GitLab's Commits API and requires no authentication to exploit. Within hours of the disclosure, security researchers began detecting active probes across the internet, with attackers already testing the vulnerability against live systems.
The vulnerability is a path traversal flaw, a class of attack that tricks an application into accessing files outside its intended directory. In this case, a malicious actor can craft a specially formed request to the Commits API that bypasses GitLab's file access controls, granting read access to sensitive data stored on the server. The simplicity of the attack—one HTTP request, no credentials needed—makes it particularly dangerous. An attacker does not need to be a GitLab user, does not need to authenticate, and does not need to know much about the target system beyond its URL.
GitLab moved quickly to alert its user base, urging immediate patching. The company understands the stakes: any organization running an unpatched instance is exposed. The files accessible through this flaw could include configuration files containing database credentials, private keys, environment variables, source code, or any other data stored on the server. For companies using GitLab to manage sensitive codebases or infrastructure, the risk of total compromise is real.
What made this disclosure particularly urgent was the speed of exploitation. Security researchers at watchTowr, among others, documented that active exploitation attempts were already underway within hours of the vulnerability becoming public. This is not a theoretical threat or a proof-of-concept that might eventually be weaponized. Attackers were already scanning the internet for vulnerable GitLab instances and attempting to extract files. The window between disclosure and widespread attack activity was measured in minutes, not days.
The detection of these probes came from multiple security firms monitoring internet traffic and attack patterns. BleepingComputer, The Hacker News, CyberScoop, and other outlets reported on the rapid exploitation activity, each documenting evidence that the vulnerability was being actively tested against real systems. This kind of immediate, widespread attack attempt is a hallmark of a flaw that is both easy to exploit and high-value to attackers.
For organizations running GitLab, the response is straightforward but urgent: patch immediately. Any delay increases the likelihood that an attacker has already accessed the system and extracted sensitive files. Beyond patching, security teams should assume that unpatched systems may have been compromised and should review access logs, monitor for data exfiltration, and rotate any credentials that might have been exposed. The vulnerability affects the Commits API specifically, so any system that exposes that API to the internet is at risk.
The disclosure also serves as a reminder of how quickly vulnerabilities can move from theoretical to actively exploited. In the time it takes to read this story, another organization somewhere is likely discovering that their GitLab instance has been probed or compromised. The race between disclosure and exploitation has become a sprint, and the only real defense is speed.
Notable Quotes
GitLab urged users to patch the maximum severity path traversal flaw immediately— GitLab (via multiple security outlets)