Defense in Depth for AI-Assisted Development

Talk Dropbox (Internal Tech Talk) February 4, 2026

Internal tech talk presenting a layered defense strategy for AI-assisted development. Covers real examples of LLM-generated security mistakes — plaintext OAuth tokens, removed CSRF protection, committed credentials — and walks through three defensive layers to catch them: pre-commit hooks with static analysis, AI-powered code review agents on pull requests, and CI workflows enforcing security scanning as a merge gate.

Based on my blog post Defense in Depth for AI-Assisted Development.

AI Security LLM DevSecOps Pre-commit Hooks CI/CD
← All Appearances View Slides