Vibecoding Promise and Peril
summary
AI code generation scales coding to millions but also creates major security gaps: empirical tests (Backspench) show models still introduce flaws and even top systems leave 20% of generated, correct code with vulnerabilities. Poor defaults (eg, Supabase misconfigurations), false safety claims (HIPAA), and crypto losses are already happening. Proposed fixes are secure-by-design opinionated platforms and a telemetry + security workflow standard so coding agents and security agents exchange plans, prompts and code.
AI-written summary · about
chapters (10)
- 0:00The promise of vibe coding
- 1:26Security risks and pitfalls
- 2:54Empirical data on model security
- 6:21Understanding the threat landscape
- 8:26Vibe coding vs. professional use
- 12:46Solutions for non-experts
- 17:20Enhancing professional workflows
- 20:16Empowering the broader security team
- 23:03Standardization and telemetry
- 26:51Future outlook and conclusions
related talks
38mVibe-Coding Your Way into a Security Nightmare
33mShipping Fast, Breaking Silently: Why AI-Generated Code Needs a Second Opinion
30mVibe Coding vs Corporate Governance: AI Innovation Without Breaking Compliance
32mFrom Vibes to Outages: Riding the AI Code Wave
23mVibes Not Vulns: Securing the Era of AI-Written Software
32mEscape the Try-Again Loop: Verifying AI-Generated Code
44mAI Code Generation - Benefits, Risks and Mitigation Controls
55mUsing GenAI on and inside your code, what could possibly go wrong?