Xbow Researchers Chain Three Zero-Days in Chrome Using AI Model

Security startup Xbow successfully chained three zero-day vulnerabilities in Google Chrome to execute code outside the browser’s sandbox on Windows machines, according to a report by The Register. The team utilized OpenAI’s GPT-5.6 Sol model to assist in the discovery and exploitation process during a six-day effort.

Google has fixed all three vulnerabilities and awarded the Xbow research team a $250,000 bounty, which is the maximum payout for a full-chain Chrome exploit. The researchers, including Brendan Dolan-Gavitt, Andres Luksenberg, Baul Lee, and Willem Pinckaers, spent approximately $40,000 on token charges to develop the attack chain during the Black Hat and DEF CON conferences.

AI-Assisted Vulnerability Research

Xbow CEO Oege de Moor stated that the speed of the discovery was more significant than the financial reward, noting that intelligence agency veterans previously took six months to achieve similar access. The team completed the entire process in just six days by leveraging the publicly available AI model.

In a technical writeup, Brendan Dolan-Gavitt explained that AI agents using the GPT-5.6-Sol model searched source code, tested hypotheses, and developed proof-of-concept stages. The human researchers provided context, challenged the AI’s results, and decided which components to chain, applying their own expertise to guide the process.

Scope and Impact

Xbow estimates that the chained vulnerabilities could affect approximately one billion users worldwide, given the prevalence of Chrome on Windows systems. The exploit bypassed Chrome’s sandbox protections, which are designed to contain compromised code and prevent attackers from taking over the entire computer through a hacked website.

Google employs multiple security layers, including sandboxes, to protect users from such threats. The successful chaining of these flaws demonstrates how AI tools can accelerate the identification of complex, multi-stage vulnerabilities that were previously difficult to discover and exploit within short timeframes.

What should you do?

Users should ensure their Chrome browsers are updated to the latest version to receive the fixes for these three vulnerabilities. Organizations should monitor for any signs of unauthorized code execution outside browser sandboxes on Windows endpoints. Security teams should review their incident response plans to address AI-assisted attack vectors and ensure rapid patch deployment processes are in place.

What to do and how to stay safe: Google Chrome

• Update Google Chrome to the latest version to apply the fixes for the three chained zero-day vulnerabilities.
• Monitor Windows endpoints for unusual processes running outside the browser sandbox that may indicate exploitation.
• Review internal policies regarding the use of AI tools in security testing to understand potential risks and benefits.
• Ensure incident response teams are prepared to handle rapid, AI-assisted attack chains that bypass traditional defenses.

Step-by-step guide: Risk-Based Vulnerability Management: Prioritize Fixes by Actual Danger

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

Frequently asked questions

How much did Xbow receive for reporting the Chrome vulnerabilities?

Google paid the Xbow team a $250,000 bounty, which is the top reward for a Chrome full-chain exploit.

What AI model did the researchers use to help find the vulnerabilities?

The team used OpenAI’s GPT-5.6 Sol model to assist in searching source code and developing exploit stages.

How many users could be affected by these vulnerabilities?

Xbow estimates that the exploit could affect approximately one billion users worldwide who use Chrome on Windows.