Phase 5: Security Testing AI-Generated Code · 55 min · Z3 SMT solver · GCC AddressSanitizer · Python

The 55% Vulnerability Problem

55.8% of AI code has Z3-proven vulnerabilities. No model passes grade D.

Real-world application: Engineers who can cite the Z3 SMT verification study and explain why 55.8% of AI code is exploitable demonstrate awareness of the scale of AI security problems.

What you will learn

Cite the Z3 SMT verification study: 55.8% vulnerability rate across 3,500 artifacts from 7 models
Explain how Z3 satisfiability witnesses provide mathematical proof of exploitability
Understand the model grading system: GPT-4o at 62.4% (grade F), Gemini 2.5 Flash at 48.4% (grade D)
Explain how 6/7 Z3-proven vulnerabilities were confirmed with GCC AddressSanitizer runtime crashes

Introduction

The 55% Vulnerability Problem

Objective

Run the Z3 SMT verification simulation on AI-generated code artifacts. Observe the 55.8% vulnerability rate, model grading, AddressSanitizer confirmation, and the 97.8% gap between Z3 and industry tools.

55.8% of AI code has Z3-proven vulnerabilities. No model passes grade D.

Unlock the full lesson

You've read the first 2 sections. The rest of this lesson covers The Problem, The Concept, Build, Check Yourself, Key Terms & Next — plus a hands-on lab, quiz, and project artifact.

Create a free account — the opening phases of 29 of 35 courses are free, no credit card.