Skip to main content

AI Security

AI security is the longer research direction of this site.

This section gathers notes on model behavior, agent systems, tool use, prompt injection, and the security questions that appear when AI is embedded into real products.

Current topics include:

  • prompt injection / jailbreak
  • tool use / agent safety
  • memory, retrieval, MCP, and browser automation risks
  • model behavior analysis
  • red teaming methods
  • AI product security review

The material here is still being sorted into a cleaner structure. Some pages are foundational; others are narrower notes tied to specific attack surfaces or system designs.