AI-assisted researchers build a Chrome exploit chain in six days, earn $250,000

[AlignmentQuarry]

Security startup Xbow used AI to help build a full Chrome exploit chain in just six days, and Google has paid it a $250,000 bounty. The Register's Jessica Lyons reports that the team used OpenAI's GPT-5.6 Sol model to help chain together three previously unknown vulnerabilities. On Windows, the chain could run code outside the browser's sandbox

Xbow estimates the flaws could have affected about one billion users. Google has fixed all three bugs, and the $250,000 payment is its top reward for a full Chrome exploit chain. Users should make sure they are running Chrome 152.0.7977.82 or later and restart the browser to apply the fix

The human team was Brendan Dolan-Gavitt, Andres Luksenberg, Baul Lee and Willem Pinckaers. They worked on it for six days during Black Hat and DEF CON, spending about $40,000 on AI token costs. AI agents searched the source code, tested ideas and built proofs of concept, while the humans picked targets, checked results and decided how to link the pieces together

The chain used a bug in V8's Maglev compiler, a flaw in the animated WebP decoder to escape the V8 sandbox, and an overflow in the Crashpad crash reporter. Chief executive Oege de Moor said a former intelligence veteran told him this kind of work used to take around six months

This is a striking example of how AI is changing security research. Finding and chaining bugs that quickly could help defenders patch faster, as it did here through responsible disclosure. It also means attackers will be using similar tools, so the speed of patching matters more than ever

Does AI assisted bug hunting help defenders more than attackers? And how quickly do you update your browser?

Security researchers spent 6 days doing their Chromework, built 1B user exploitwww.theregister.com