Vibes Not Vulns: Securing the Era of AI-Written Software
summary
AI-assisted development already pushes production apps out in hours with Copilot, Claude, Cursor and autonomous agents, and research from Aikido Security finds 60% of orgs see AI-related incidents and 20% see serious ones. A 1,000-prompt study across five languages shows many functions compile with vulnerabilities unless prompted “make sure it’s secure.” CI/CD use of AI yields prompt injection leading to RCE and leaked keys (Gemini CLI case), dependencies suffer hallucination squatting, and large npm compromise hit 2.6 billion weekly downloads and infected ~10% of cloud accounts within hours.
AI-written summary · about
chapters (7)
- 0:00AI coding security risks
- 1:28Real-world vulnerabilities
- 6:23Prompt injection threats
- 10:35Prompt injection in CI/CD
- 15:55Supply chain vulnerabilities
- 20:55AI malware detection
- 22:40AI vulnerability triage
related talks
12mTrusting AI with Code: How Secure is AI-written code?
33mShipping Fast, Breaking Silently: Why AI-Generated Code Needs a Second Opinion
26mSmarter Code, Sneakier Risks: Supply Chain Security in the Age of AI
31mHacking AI Applications
58mSecuring Code in the Age of AI
59mUsing AI for Secure Code Creation: Enhancing Software Security
45m📙 Prompt, Commit, Repeat: Security at Scale When 1,000 Devs Go AI-Native
38mVibe-Coding Your Way into a Security Nightmare