Human vs AI: How to Ship Secure Code
summary
GitHub Security Lab reports 300–400 zero days found over three years — roughly one every three days — and research shows one application security specialist per 100 developers. Demonstrations use GitHub Copilot, Copilot Labs brushes and Copilot Chat (coming to GA), and Copilot X vision; claim that roughly 46% of committed code is AI-driven. AI generates fixes, tests and CodeQL queries quickly and a free coding game exists, but it misses attack‑surface discovery and fuzzing entry points, so generated code still requires manual testing and harnessed SAST tools.
AI-written summary · about
chapters (10)
- 0:00Introduction and background
- 2:14Why cybersecurity is hard
- 3:21AI's role in security gaps
- 7:21Privacy and AI code generation
- 8:43Writing safer code with AI
- 15:52Testing AI-generated code
- 16:56Finding bugs with AI assistance
- 19:24Developer training with AI
- 22:11AI's limitations in security
- 26:55Summary and next steps
related talks
41mCode Security Reinvented Navigating the era of AI
58mSecuring Code in the Age of AI
38mCode Security Reinvented: Navigating the era of AI
32mCode Security Reinvented: Navigating the Era of AI
23mAI-Generated Code: Who Is Responsible for the Bugs?
01h 02mHow AI is impacting cybersecurity. Does it?
48mAI Impact on Cybersecurity and Privacy
15mThe reality of developers using AI | State of Code Developer Survey