Chrome Sandbox Escape: AI-Assisted Chain of Three Zero-Days Impacts Windows Users

Xbow researchers achieved full system access on Windows machines by chaining three zero-day vulnerabilities in Google Chrome, according to The Register. The exploit allowed code execution outside the browser’s sandbox, bypassing the security barriers designed to contain compromised web content. This capability gives an attacker total control over the host system, turning a browser compromise into a complete machine takeover.

How it unfolded

• Xbow researchers spent six days during Black Hat and DEF CON conferences building the exploit chain.
• The team used OpenAI’s GPT-5.6 Sol model to search source code and develop proof-of-concept stages.
• Human researchers selected targets, provided context, and validated the AI-generated results.
• Google issued patches for all three flaws and paid a $250,000 bounty to the four researchers.

Who is affected

The vulnerability chain impacts approximately one billion Chrome users worldwide, according to Xbow estimates. The exploit specifically targets Windows machines, allowing attackers to break out of the browser’s sandbox environment. This means that any user running Chrome on Windows could potentially be affected if they encounter the specific attack vector used in this demonstration.

The fix

Google has released patches for all three vulnerabilities involved in the exploit chain. The company paid the top-tier $250,000 bounty for this full-chain report. Users should update their Chrome browsers to the latest version to ensure they have these fixes applied. No further action is required beyond installing the standard browser updates provided by Google.

What made it possible

Sandbox escape vulnerabilities typically arise when a browser process fails to strictly enforce memory or execution boundaries. Attackers exploit these weaknesses to move from a restricted web context to the host operating system. Managed service providers should monitor for alerts indicating processes spawning outside expected browser containers. Teams usually analyze telemetry for unusual privilege escalations that follow web activity. Checking whether endpoint detection tools flag anomalous process creation can help identify these lateral movement attempts early.

What to do and how to stay safe: Google Chrome

• Ensure all client machines running Chrome on Windows are updated to the latest stable version.
• Monitor endpoint logs for processes that spawn unexpectedly after web browsing activity.
• Verify that browser sandbox integrity checks are enabled and functioning correctly.
• Review access controls to limit the damage potential of any single compromised browser session.

Step-by-step guide: What Is Virtual Patching: How It Works Without Code Changes

General security guidance from the RedTeam Tribune newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

How many users are affected by this Chrome exploit?

Xbow estimates that approximately one billion users worldwide could be affected by this vulnerability chain.

What role did AI play in creating this exploit?

The researchers used OpenAI’s GPT-5.6 Sol model to search source code, test hypotheses, and develop exploit stages.

Has Google released a fix for these vulnerabilities?

Yes, Google has patched all three vulnerabilities and paid a $250,000 bounty for the report.