A security research team spent six days developing a complete exploit chain for Google Chrome using an open-source AI model, ultimately triggering a $250,000 bug bounty from Google. The researchers, affiliated with the startup Xbow, utilized OpenAI's GPT-5.6-Sol model to identify and chain three distinct zero-day vulnerabilities in Chrome. These vulnerabilities, when combined, enabled execution of code outside the browser's sandbox on Windows systems, potentially impacting approximately one billion users globally. Google addressed all three flaws and awarded the maximum bounty for a full-chain Chrome exploit. The team, comprising Brendan Dolan-Gavitt, Andres Luksenberg, Baul Lee, and Willem Pinckaers, reported spending roughly $40,000 on computational resources during the Black Hat and DEF CON conferences to develop the exploit. Xbow CEO Oege de Moor noted that the speed of their discovery, facilitated by AI assistance, contrasted sharply with historical timelines where similar vulnerabilities might take security agencies six months to uncover. In a technical analysis, Dolan-Gavitt described how AI agents processed source code, tested hypotheses, and generated exploit components, while human researchers curated targets, validated results, and determined the attack sequence. The exploit chain bypassed Chrome's layered security architecture, including the V8 JavaScript engine sandbox and nested process restrictions. It leveraged a compiler bug in V8's Maglev optimizer, a flaw in the animated WebP decoder, and an overflow vulnerability in the Crashpad crash reporting tool. Google confirmed patching all issues in Chrome version 152.0.7977.82 or later, though fixes require browser restarts to activate. The case underscores growing concerns about how artificial intelligence tools may lower the barrier for sophisticated cyberattacks, prompting calls for enhanced defensive measures across the industry.