Project Zero: BigSleep beat Fuzzilli on V8 bugs, overlap partial
Original title: Fuzzers vs. AI: Comparing Fuzzilli and BigSleep on JavaScript Engines
In six months BigSleep reported 45 V8 vulnerabilities to Fuzzilli's 37, and experiments put theoretical overlap near 60%.
Google Project Zero compared Fuzzilli, a generative JavaScript fuzzer, with BigSleep, an LLM-based bug finder, on Chrome's V8 engine. From April through September 2026, BigSleep found more vulnerabilities than Fuzzilli, 45 versus 37, with 25% overlap in the specific bugs reported. Dedicated experiments on vulnerabilities reported since 2023 estimated overlap at roughly 60%. The team says many bugs were found by only one method and recommends vendors run both fuzzing and agentic analysis.
Unconfirmed: Reported-bug overlap is 25%, while a separate experiment estimates about 60%; the excerpt does not give final actionable totals.
- Unfiltered reports totaled 316 for Fuzzilli and 441 for BigSleep.
- Issues typed as vulnerabilities were 178 for Fuzzilli and 95 for BigSleep.
- Fuzzilli runs continuously; BigSleep reviews new commits about once a day.
- Both systems file reports through ClusterFuzz into Chrome's issue tracker.
- The excerpt ends before final counts of actionable, fixed vulnerabilities.
- 1.Run both fuzzing and agentic code analysis to maximize vulnerabilities found before users are affected.
- Vendors
- Organizations
- Project Zero
From the original article
14 min read at projectzero.googleSummary Project Zero compared two automated tools used by Google to find vulnerabilities in Chrome’s JavaScript engine, V8, on a regular basis. Fuzzilli is a generative JavaScript engine fuzzer and BigSleep is an LLM-based bug finding system. In the past six months, BigSleep found more vulnerabilities than Fuzzilli (45 versus 37) with a 25% overlap in the specific bugs reported.
By Samuel Groß · projectzero.google
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.