← Back to Feed CACHED · 2026-05-17 09:42:19 · cache_key CVE-2025-29912
CVE-2026-10914 · CWE-416 · Disclosed 2026-06-04

Use after free in ANGLE in Google Chrome on Windows prior to 149

ASSESSED — NOISGATE V0.5
Vendor
Reassessed
Verdict:
01 · The Real Story

This is a burglar getting into the apartment lobby, not the penthouse safe

CVE-2026-10914 is a use-after-free in ANGLE affecting Google Chrome on Windows before 149.0.7827.53. In plain English: a malicious page can corrupt memory in graphics-handling code and potentially get attacker-controlled code running inside Chrome's sandboxed process, not directly as the user on the underlying OS. The patch line landed in Chrome 149 stable on June 2, 2026, while the CVE/public metadata appears to have surfaced on June 4, 2026.

Google's HIGH/8.8 label is technically defensible for a memory-corruption bug in a massively deployed browser, but it overstates the enterprise patch urgency when assessed as a standalone bug. The decisive friction is that the stated impact is sandboxed code execution on a client browser, which usually means the attacker still needs a second bug for sandbox escape, token theft beyond the browser boundary, or meaningful host compromise; add user interaction, Windows-only scope, no KEV entry, and an EPSS of 0.0008, and this drops out of urgent-fire territory into solid MEDIUM.

"Sandboxed renderer code exec is bad, but this is still one stage short of the host takeover everyone fears."
02 · The Attack Path

4 steps from start to impact.

STEP 01

Land the user on a hostile page

The attacker needs a delivery vehicle: phishing, malvertising, SEO poisoning, or a compromised site. The page must execute browser-accessible content that reaches the vulnerable ANGLE path, likely through graphics-heavy HTML/JavaScript or related rendering behavior rather than a raw network packet exploit.
Conditions required:
  • Victim uses Google Chrome on Windows
  • Version is < 149.0.7827.53
  • Victim browses attacker-controlled or attacker-influenced content
Where this breaks in practice:
  • Requires user interaction (UI:R)
  • Enterprise mail filtering, Safe Browsing, DNS filtering, and ad blocking cut a lot of commodity delivery
  • Non-Windows Chrome population is irrelevant to this CVE as described
Detection/coverage: Scanner coverage is mostly version-based; network IDS will rarely see exploit semantics here. Browser extension telemetry, proxy logs, and Safe Browsing events are more realistic than signature matching.
STEP 02

Trigger memory corruption in ANGLE

The hostile page drives Chrome into the vulnerable ANGLE code path and attempts to hit the use-after-free reliably enough for controlled execution. In practice this is the hard engineering part: modern browser exploitation needs heap shaping, version matching, and stability work against a noisy client environment.
Conditions required:
  • Reachable vulnerable ANGLE code path
  • Exploit tuned for the victim's Chrome build and Windows environment
Where this breaks in practice:
  • Chrome's exploit mitigations, sandboxing, CFG/ACG-class platform defenses, and version drift make reliable weaponization non-trivial
  • Bug details are typically restricted at release time, slowing opportunistic copycat exploitation
Detection/coverage: EDR may catch follow-on crash spikes, suspicious child-process behavior, or exploit mitigation events, but commodity vuln scanners will not validate exploitability.
STEP 03

Gain code execution in the sandboxed process

If exploitation works, the attacker gets code running inside the sandbox associated with the compromised Chrome process, not unrestricted host execution. That can still expose session context, page content, or create a foothold for chaining, but it is materially less than a browser-process or OS-level compromise.
Conditions required:
  • Successful memory corruption exploitation
  • Chrome process remains constrained by sandbox policy
Where this breaks in practice:
  • Sandbox boundaries sharply limit direct file system, registry, device, and broader OS access
  • Many enterprise attack objectives still require credential theft outside the renderer or a second-stage escape
Detection/coverage: Crash telemetry, renderer instability, unusual GPU/renderer behavior, and EDR detections around exploit chains are the best clues; pure version scanners stop at 'vulnerable'.
STEP 04

Chain to real host impact

To turn this into the kind of incident that wakes up the SOC, the attacker usually needs a sandbox escape, broker abuse, local privilege escalation, or data theft objective confined to browser context. That is the key severity brake: this CVE is dangerous, but by itself it is often an exploit-chain component rather than the full compromise story.
Conditions required:
  • A second vulnerability or a browser-resident data objective
  • Sufficient post-exploitation path from sandbox to mission impact
Where this breaks in practice:
  • No public evidence here of an accompanying escape or active campaign
  • Modern EDR, attack surface reduction, and rapid browser auto-update often break the chain before host compromise
Detection/coverage: Look for pairing with Windows EoP, broker abuse, suspicious downloads, or post-browser execution. Alone, this bug's most realistic enterprise detection remains asset/version hygiene.
03 · Intelligence Metadata

The supporting signals.

In-the-wild statusNo public active exploitation evidence found. Not present in CISA KEV based on the current catalog.
Proof-of-concept availabilityNo public PoC located in primary-source review. Chromium bug details remain gated/restricted via the linked issue tracker, which usually slows opportunistic weaponization.
EPSS0.0008 from the provided intel block; that is extremely low near-term exploitation pressure. Percentile was not provided.
KEV statusNot KEV-listed. That matters: if this were showing broad in-the-wild abuse, KEV would be a strong upward pressure.
CVSS vector readoutCVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H = remotely triggerable with no auth, but user interaction is required and scope stays unchanged; the vendor score assumes the sandboxed code-exec impact is still severe.
Affected versionsGoogle Chrome on Windows prior to 149.0.7827.53 per the intel you supplied; Chrome 149 stable shipped on 2026-06-02.
Fixed version149.0.7827.53 or later on Windows. The desktop stable rollout announced 149.0.7827.53/54 for Windows/Mac and 149.0.7827.53 for Linux in the Chrome release post.
Scanning / exposure dataInternet exposure platforms are the wrong lens. This is endpoint client software, not a server you can enumerate via Shodan/Censys/FOFA; exposure is governed by Windows Chrome fleet versioning, not public attack surface.
Disclosure timingPatch release: 2026-06-02. CVE/public metadata supplied: 2026-06-04. That two-date split is normal for Chrome.
Reporter / sourceThe Chrome release entry attributes this CVE as reported by Google on 2026-03-30 via Chromium issue 497574371.
04 · The Call

noisgate verdict.

Final Verdict
DOWNGRADED to MEDIUM (6.3/10)

The single biggest downgrade driver is impact containment by the Chrome sandbox. This bug may get an attacker code execution in a renderer-adjacent process, but without an accompanying escape or broker abuse path it is usually not yet a host compromise, and that materially changes patch priority for a 10,000-endpoint fleet.

HIGH Affected/fixed version mapping
MEDIUM Standalone exploitability assessment
MEDIUM Real-world enterprise impact without a chained escape

Why this verdict

  • Sandboxed code exec cuts the blast radius: vendor 8.8 starts from memory corruption in a major browser, but the stated impact is execution inside a sandbox, not unrestricted code execution on the endpoint. That is a meaningful downward adjustment from host-takeover territory.
  • User interaction narrows reach: the attacker needs a user to hit a malicious page on vulnerable Windows Chrome. Web filtering, Safe Browsing, mail protections, and ordinary user behavior reduce the exploitable population compared with unauthenticated server-side bugs.
  • Threat telemetry is cold: no KEV listing, no public exploitation evidence found, and EPSS 0.0008 all argue against treating this as an emergency patch event despite the scary vulnerability class.

Why not higher?

It is not higher because the public description stops at sandboxed execution. For browser bugs, the line between 'serious' and 'drop everything' is usually whether the attacker gets out of the sandbox or whether there is confirmed active exploitation; neither is present here from the available evidence.

Why not lower?

It is not lower because this is still memory corruption in a ubiquitous browser, delivered over normal browsing with no authentication and potentially exploitable by a malicious page. If you run a large Windows fleet, even sandboxed browser code execution deserves disciplined version compliance, because it can be paired with other bugs or used for browser-scoped data theft.

05 · Compensating Control

What to do — in priority order.

  1. Enforce browser auto-update — Make sure managed Windows endpoints cannot drift below 149.0.7827.53. For a MEDIUM verdict there is no mitigation SLA — go straight to the 365-day remediation window, but in practice browser drift should be cleaned up far sooner because update enforcement is cheap and high leverage.
  2. Reduce untrusted web execution paths — Use enterprise policy, web filtering, ad blocking, and Safe Browsing controls to lower the odds of users reaching exploit delivery pages. There is no mitigation SLA for MEDIUM here, so apply this as a risk-reduction measure where patch latency or kiosk exceptions exist while staying inside the 365-day remediation window.
  3. Watch for crash and exploit telemetry — Tune EDR and browser telemetry for unusual Chrome renderer/GPU crashes, exploit mitigation events, and suspicious child-process or follow-on activity. That will not prove this CVE specifically, but it is the most practical detection layer if a browser memory-corruption chain is attempted during the remediation window.
  4. Isolate exception systems — For VDI images, kiosks, lab systems, or pinned application stacks that cannot move immediately, tighten egress and browsing scope to approved destinations only. Again, no mitigation SLA applies at MEDIUM, but exception populations should still be documented and remediated within the 365-day patch window.
What doesn't work
  • WAFs do not meaningfully help; this is a client-side browser bug reached through normal page rendering, not a server request you can sanitize at your perimeter.
  • MFA does nothing for initial exploitation of the browser memory corruption itself; it only helps with downstream account abuse if credentials or sessions are later targeted.
  • Generic network vuln scanning will not validate exploitability. At best it tells you installed version; it will not tell you whether the vulnerable ANGLE path is reachable or weaponized.
06 · Verification

Crowdsourced verification payload.

Run this on the target Windows endpoint or through your endpoint management shell. Invoke it with powershell -ExecutionPolicy Bypass -File .\check-chrome-cve-2026-10914.ps1; standard user rights are usually enough because it only reads file version and uninstall metadata.

noisgate-verify.ps1
POWERSHELLREAD-ONLYSAFE
# check-chrome-cve-2026-10914.ps1

# Purpose: Detect whether Google Chrome on Windows is vulnerable to CVE-2026-10914

# Logic: Vulnerable if installed Chrome version is less than 149.0.7827.53

# Exit codes: 0=PATCHED, 1=VULNERABLE, 2=UNKNOWN


$ErrorActionPreference = 'Stop'

function Get-VersionObject {
    param([string]$VersionString)
    try {
        return [version]$VersionString
    } catch {
        return $null
    }
}

function Get-ChromeVersion {
    $paths = @(
        "$env:ProgramFiles\Google\Chrome\Application\chrome.exe",
        "$env:ProgramFiles(x86)\Google\Chrome\Application\chrome.exe",
        "$env:LocalAppData\Google\Chrome\Application\chrome.exe"
    )

    foreach ($p in $paths) {
        if (Test-Path $p) {
            try {
                $fv = (Get-Item $p).VersionInfo.ProductVersion
                if ($fv) {
                    return [pscustomobject]@{ Path = $p; Version = $fv; Source = 'Executable' }
                }
            } catch {}
        }
    }

    $regPaths = @(
        'HKLM:\SOFTWARE\Microsoft\Windows\CurrentVersion\Uninstall\*',
        'HKLM:\SOFTWARE\WOW6432Node\Microsoft\Windows\CurrentVersion\Uninstall\*',
        'HKCU:\SOFTWARE\Microsoft\Windows\CurrentVersion\Uninstall\*'
    )

    foreach ($rp in $regPaths) {
        try {
            $apps = Get-ItemProperty $rp -ErrorAction SilentlyContinue |
                Where-Object { $_.DisplayName -match '^Google Chrome$' -and $_.DisplayVersion }
            foreach ($app in $apps) {
                return [pscustomobject]@{ Path = 'Registry'; Version = $app.DisplayVersion; Source = 'UninstallKey' }
            }
        } catch {}
    }

    return $null
}

$fixedVersion = Get-VersionObject '149.0.7827.53'
$found = Get-ChromeVersion

if (-not $found) {
    Write-Output 'UNKNOWN - Google Chrome not found on this host or version could not be determined.'
    exit 2
}

$currentVersion = Get-VersionObject $found.Version
if (-not $currentVersion) {
    Write-Output ("UNKNOWN - Found Chrome via {0}, but version '{1}' could not be parsed." -f $found.Source, $found.Version)
    exit 2
}

if ($currentVersion -lt $fixedVersion) {
    Write-Output ("VULNERABLE - Google Chrome version {0} detected via {1}; fixed version is {2}." -f $currentVersion, $found.Source, $fixedVersion)
    exit 1
} else {
    Write-Output ("PATCHED - Google Chrome version {0} detected via {1}; meets or exceeds fixed version {2}." -f $currentVersion, $found.Source, $fixedVersion)
    exit 0
}
07 · Bottom Line

If you remember one thing.

TL;DR
Monday morning, treat this as version-compliance work, not incident-response work: query your Windows fleet for Chrome versions, verify update policy is actually moving endpoints to 149.0.7827.53+, and carve out kiosks/VDI/gold images that are stuck below that floor. For this MEDIUM verdict there is no noisgate mitigation SLA — go straight to the 365-day remediation window; in other words, don't invent emergency controls, but do close version drift and complete patching within the noisgate remediation SLA of ≤365 days.

Sources

  1. Chrome Releases - June 2026 stable desktop update
  2. Chromium issue 497574371
  3. Chromium severity guidelines
  4. Chrome sandbox diagnostics for Windows
  5. Chromium memory safety overview
  6. CISA Known Exploited Vulnerabilities catalog
  7. FIRST EPSS API documentation
  8. Chromium security release management
Peer Review

What defenders are saying.

Submit a review attribution: handle + country only
0 flags selected · stored anonymously
Validation Results

Crowdsourced verification outputs.

Results submitted by users who ran the verification payload against their environment.