Skip to content
AI|Course

Making systems secure and reliable · Intermediate

Analyze AI-Generated Code for Quality and Security

AI tools write plausible code that can be wrong, insecure or built on packages that do not exist. Learn to review it by risk, recognize common weaknesses, vet dependencies, automate analysis and test it properly.

7 lessons · 15 hr · Free

You will learn to

  • ▸Explain what research shows about the quality and security of AI-generated code
  • ▸Review AI-written changes by risk, with a checklist for known failure patterns
  • ▸Recognize and fix the weaknesses AI tools reproduce most often
  • ▸Vet dependencies, defend against hallucinated packages and generate an SBOM
  • ▸Build a layered CI pipeline and tests that genuinely protect you, and measure it

Before you start

  • ·You write code and use pull requests
  • ·Basic familiarity with Python or JavaScript, since examples use Python
  • ·You have used an AI coding assistant

Syllabus

  1. 01Why AI-generated code needs a different reviewWhat the research found, and five ways AI code goes wrong.1 hr
  2. 02Reviewing by riskIntent first, risk map, known patterns and running the code.1 hr 30 min
  3. 03Common security weaknessesInjection, traversal, randomness, secrets and access control, with fixes.2 hr
  4. 04Dependencies and the supply chainHallucinated packages, vulnerabilities, SBOMs, SLSA and licenses.1 hr 30 min
  5. 05Automated analysis and governanceLinters, types, SAST, secret scanning, LLM reviewers and policy.2 hr
  6. 06Testing AI-generated codeProperty-based tests, mutation testing, fuzzing and safe execution.2 hr
  7. 07Capstone projectBuild a pipeline that gates AI-assisted changes, and measure it.5 hr