Google Chrome · Public exploit code

Chrome 156 Patches 52 Flaws; Anthropic AI Escaped Sandbox

Oct 19, 2026 · 1 min read

Google shipped stable Chrome 156 for desktop with fixes for 52 security flaws, 4 of them critical, according to SecurityWeek and TheHackersNews. A public writeup details that an unreleased Anthropic AI model built five working exploit chains across Chrome, Safari, Android, Windows, and Linux; in Chrome, one escaped the sandbox and launched Calculator on macOS just by loading a page. Google makes no mention of exploitation in the wild; update to 156.0.8078.47/.48 (Windows and Mac) or 156.0.8078.47 (Linux).

By HIFY Radar newsroom Updated 10/28/2026 10:44 AM Get this every day →← Back to the feed 3 sources

HighVulnerabilityPublic exploit codePatched

In this article

What happened
Why it matters
What to do
For the technical team
Timeline
Sources

What happened

Google released Chrome 156 for desktop with fixes for 52 security flaws, 4 of them critical, according to SecurityWeek. A public writeup reports that an unreleased Anthropic AI model built five working exploit chains, including one that escaped Chrome's sandbox. Google has not mentioned exploitation in the wild.

Why it matters

One critical flaw, CVE-2026-118306, allows code execution outside the sandbox through a crafted HTML page, according to SecurityWeek. Any employee on an outdated Chrome can be compromised by opening a link. The case also shows AI models can already find and chain real flaws in widely used products.

What to do

Update Chrome to 156.0.8078.47/.48 (Windows and Mac) or 156.0.8078.47 (Linux) on all company computers. Force the update via management policy and verify installed versions.

For the technical team

Affected: Google Chrome (desktop) · Google Chrome (desktop) before 156.0.8078.47 (Google Chrome Releases).
Exploitation: No exploitation in the wild reported; Google and Mozilla make no mention of it, according to SecurityWeek. A public writeup documents five working exploit chains built by an unreleased Anthropic AI model (TheHackersNews).
Fix: Chrome 156.0.8078.47/.48 (Windows and Mac) and 156.0.8078.47 (Linux), per Google Chrome Releases; SecurityWeek cites 156.0.8078.73/.74 (Windows and macOS) and 156.0.8078.73 (Linux).
Vector: Desktop browser. Critical flaws: memory corruption in FileSystem, improper input validation in FileAPI and Skia, and use-after-free in Layout (SecurityWeek). CVE-2026-118306 (FileSystem) allows arbitrary code execution outside the sandbox through a crafted HTML page. The AI chain fired just by loading a page (TheHackersNews).
More details: According to TheHackersNews, the AI Claude discovered dozens of Chrome vulnerabilities and built a working multi-stage exploit chain that escaped the browser sandbox and executed code. SecurityWeek counts 46 high-severity fixes and reports no reward for 39 of the 52 flaws, most found internally or with AI assistance. On the same day, Mozilla released Firefox 159 with 167 fixes (108 high-severity), plus Thunderbird 159 and 153.6 and Firefox ESR 153.6 and 115.44. Rollout start: 2026-10-20 (Google Chrome Releases).

Timeline

10/28/2026 10:44 AM
Roundup coverage tallies 52 security flaws fixed in the Chrome 156 update.
10/28/2026 05:18 AM
Public writeup details five exploit chains built by an unreleased Anthropic AI model; the Chrome chain escaped the sandbox and launched Calculator on macOS from a single page load.
10/26/2026 01:20 PM
Incident opened.