Skip to main content

AI Infrastructure Security Game

Interactive AI infrastructure security simulator: identify indirect prompt injection, model extraction, vector-DB tenant leakage, agent tool over-scoping, inference cost abuse, and MCP server identity gaps across 6 production scenarios. Free, no signup.

Interactive AI security lab. Six scenarios cover indirect prompt injection in support agents and tool-layer capability scoping, model extraction via high-volume diverse queries and the multi-layer defence stack, multi-tenant vector DB isolation in shared indexes, agent tool over-scoping and capability-based decomposition, free-tier inference API abuse with disposable email farming, and MCP server identity propagation via SPIFFE workload identity.

What You Practice

  • Recognizing production failure modes before they become incidents.
  • Connecting security, reliability, and operational choices to real engineering outcomes.
  • Building intuition through short interactive scenarios instead of passive reading only.

Related Practice

Continue with free courses or read production engineering articles.