How Much Trust Do You Have with LLM-Based Solutions?
summary
Claims that developer AI yields ~50–55% efficiency gains but 40% of AI-generated lines have security flaws and roughly 80% of existing codebases contain critical flaws. Observes remediation lag: only ~60% of vulnerabilities are fixed after a year so 40% persist. Describes a built solution: an IDE-integrated scanner/remediator in VS Code that trains on a curated security dataset rather than open source and aims to cut time-to-fix from months to minutes, preventing many vulnerabilities at commit time.
AI-written summary · about
chapters (6)
- 0:00Introduction
- 0:43How much do you trust AI
- 3:42The problem
- 4:40The solution
- 6:20Scaling forward
- 7:50Closing thoughts
related talks
10mAI Coding & Cybersecurity — Why AI coding tools aren’t ready for production yet
33mShipping Fast, Breaking Silently: Why AI-Generated Code Needs a Second Opinion
57mKeynote: Using LLMs and generative AI to fix software vulnerabilities
13mIntelligence in AppSec: Use AI to supercharge DevSecOps
12mTrusting AI with Code: How Secure is AI-written code?
48mAI Impact on Cybersecurity and Privacy
32mCan We Trust AI-Generated Code?
33mAccelerate Code Quality in the AI Age