Open Source AI reality
summary
Open source projects see a shift from noisy hallucinations in 2025 to high-volume, high-quality reports in 2026: report rate doubled in 2025 and doubled again, now averaging one report per day (roughly 20 hours per report vs ~110 before). Modern AI-powered code analyzers read uncompiled variants, cross libraries and protocols, generate reproducers and patch suggestions, and surface decades‑old bugs; curl prepares 8 CVEs in next curl release (18 CVEs so far in 2026). Fixes still need human review and triage.
AI-written summary · about
chapters (5)
- 0:00AI security reality
- 3:46Generative AI awakening 2025
- 8:46From AI slop to quality reports
- 27:50Security reports drowning era
- 38:47Q&A
related talks
28mScalable Security and Compliance in the Age of AI
19mIf AI Writes the Code, Who Runs Production?
11mAIxCC Results and New Open Source AI Projects To Help Secure Open Sou...
10mThe Next Era Of Open Source: Humans + AI, Not Humans vs AI
20mThe AI bugpocalypse is here. Now what?
24mAI Is Writing Your Bug Reports. Can You Tell?
31mAI for Web3 Security
43mWriting for Machines: How To Capture Your Project's "Vibe" and Survive AI Slop