AIxCC Results and New Open Source AI Projects To Help Secure Open Sou...
summary
DARPA and ARPA-H’s two-year AIxCC competition produces seven open-source cyber reasoning systems; finals scan 54+ million lines, find 86% of synthetic bugs and patch 68%, discover 18 real-world zero-days and land 11 patches. Typical patch cost averages ~$152 per task, with some fixes at one line and one at 300+ lines. OpenSSF coordinates human-reviewed issue reports, partners with OTF to approach projects, offers small stipends to maintainers, and publishes the systems and data as open source.
AI-written summary · about
chapters (9)
- 0:00AI Cyber Challenge intro
- 1:05Challenge overview and goals
- 2:11OpenSSF role
- 3:15Competition results
- 5:00Finals performance
- 6:48Next steps for adoption
- 8:15Human-centered AI security
- 9:17Engaging projects with findings
- 10:19Post-competition collaboration
related talks
44mArtificial Intelligence Cyber Challenge (AIxCC): Overview and Releasing Research as Open Source Software
53mCracking the Code: Unveiling Synergies Between Open Source Security and AI
01h 02mPanel Discussion:The Present and Future of AI and Security
21mNavigating the Intersection: AI’s Role in Shaping the Secure Open Source Software Ecos...
15mOSS-CRS: Liberating AIxCC Cyber Reasoning Systems for Real-World Open-Source Security
31mAI for Web3 Security
14mAI's impact on open-source contributions and mentoring
41mSkill Degradation: An Empirical Analysis of 400+ AI‑Generated Security Fixes