Tracked
5,131
Commit mapped
4935
96% of tracked
Embargoed / no public fix
196
the live patch gap
Exploited in the wild
13
Root-cause analysis
2419
49% of mapped
Bug report / PoC
3083 / 1002

Exposure window — days a fix sat in public source before Chrome told users

For each CVE: the date the fixing commit landed in public git, minus the date of Chrome's security advisory. During that window anyone reading the commit log could see the fix, while most users had not yet been told to update. Built only from public, checkable timestamps (git commit date, advisory date, the advisory's "Reported by … on date" line). No AI analysis feeds these numbers.
High/Critical memory-safety bugs
15 days
Half of these fixes were public for 15+ days before the advisory. Middle half: 8–33 days. 27.1% (254) were public for over 30 days. n = 937.
Exploited in the wild
5 days
Median for the 13 CVEs the advisory or CISA KEV marks as exploited. Middle half: 2.5–21 days. Each one is listed below.
All CVEs (any severity)
35 days
Middle half: 18–44 days. 62.7% over 30 days. n = 3,045. Low/Medium fixes are often held for a later release, which pulls this up.
Fix shipped a regression test
64.6%
Of all dated fixes. Median 35 days public before the advisory; 1262 for 30+ days. A test in the commit often encodes the trigger.

Lifecycle (median, 3,039 CVEs with a report date): reported → fix landed 12 days; fix landed → advisory 35 days; reported → advisory 50 days.

Exposure by component — High/Critical memory-safety bugs

Median days public before the advisory, for components with at least 8 such bugs. Hover for the count and how many exceeded 30 days.

Exploited in the wild — exposure per CVE

CVEVendor titleFix landedAdvisoryDays
CVE-2026-11645 Out of bounds memory access in V8 2026-05-04 2026-06-08 35
CVE-2026-87491 Out of bounds write in V8 2026-08-11 2026-09-08 28
CVE-2026-85046 Type confusion in V8 2026-08-07 2026-09-03 27
CVE-2026-5281 Use after free in Dawn 2026-03-16 2026-03-31 15
CVE-2025-6558 Incorrect validation of untrusted input in ANGLE and GPU 2025-07-02 2025-07-15 13
CVE-2025-5419 Out of bounds read and write in V8 2025-05-27 2025-06-02 6
CVE-2025-13223 Type Confusion in V8 2025-11-12 2025-11-17 5
CVE-2025-6554 Type Confusion in V8 2025-06-26 2025-06-30 4
CVE-2025-24201 Out of bounds write in GPU 2025-03-07 2025-03-10 3
CVE-2026-3909 Out of bounds write in Skia 2026-03-10 2026-03-13 3
CVE-2026-3910 Inappropriate implementation in V8 2026-03-10 2026-03-12 2
CVE-2025-10585 Type Confusion in V8 2025-09-16 2025-09-17 1
CVE-2026-2441 Use after free in CSS 2026-02-12 2026-02-13 1

Longest exposure — High/Critical memory-safety bugs

CVEVendor titleSeverityFix landedAdvisoryDays
CVE-2026-8532 Integer overflow in XML upstream fix high 2025-10-03 2026-05-12 221
CVE-2026-87578 Use after free in Receiver high 2026-05-30 2026-09-08 101
CVE-2026-93374 Use after free in Dawn critical 2026-06-22 2026-09-17 87
CVE-2026-3919 Use after free in Extensions high 2025-12-22 2026-03-10 78
CVE-2026-84353 Use after free in Shared Tab Groups critical 2026-06-24 2026-09-01 69
CVE-2026-17667 Uninitialized Use in ANGLE high 2026-05-22 2026-07-29 68
CVE-2026-9113 Out of bounds read in GPU high 2026-03-12 2026-05-19 68
CVE-2026-2313 Use after free in CSS high 2025-12-10 2026-02-10 62
CVE-2026-14385 Heap buffer overflow in ANGLE high 2026-05-04 2026-06-30 57
CVE-2025-0434 Out of bounds memory access in V8 high 2024-11-19 2025-01-14 56
CVE-2026-13803 Type Confusion in Chrome Tabs high 2026-05-05 2026-06-30 56
CVE-2026-17652 Use after free in Views critical 2026-06-03 2026-07-29 56
CVE-2026-17682 Integer overflow in ANGLE high 2026-06-03 2026-07-29 56
CVE-2026-17697 Type Confusion in ANGLE high 2026-06-03 2026-07-29 56
CVE-2026-17704 Use after free in ANGLE high 2026-06-03 2026-07-29 56

Definitions. High/Critical memory-safety = Google rated it High or Critical and the advisory title names a memory-safety bug (use after free, out of bounds, heap/stack buffer overflow, type confusion, integer overflow, uninitialized use, double free). It uses only the vendor's own labels, not PatchGap's AI tiers. Exploited in the wild = the advisory says an exploit exists in the wild, or CISA KEV lists the CVE. Middle half = 25th to 75th percentile. For third-party code (HarfBuzz, libexpat, …) the fix date is the upstream commit, which can be public long before Chrome rolls it in; those rows are marked "upstream fix". Method. 22 CVEs whose mapped commit is dated after the advisory (typically a later merge or backport) are excluded from medians. A further 36 are left out because their commit mapping is flagged as unverified, or because the only public commits citing the bug are measurement, test or documentation changes (their dates say nothing about when the code fix became public). Very long windows are where mapping errors concentrate, so check the linked commit before quoting an outlier. Data generated 2026-11-02.

Patch gap over time — mapped vs. embargoed per advisory month

By severity

By vendor

Top components

Top bug classes

Analysis coverage

Root-cause analysis2419 / 5131
Original bug report linked3083 / 5131
Reporter PoC files1002 / 5131

Root-cause analysis is written only for CVEs whose fixing commit is public, so coverage is measured against mapped CVEs on the cards above.

Patch gap — fix availability

Fixing commit located4935
Embargoed / not yet public196

A CVE counts as mapped once its fixing change is public in the vendor repository. The remainder is the live gap between disclosure and a locatable public fix.