MSP 2026 • AI HACKING COMPETITION

MSP • AI CTF

Realistic AI Application Security Failures, Direct & Indirect Prompt Injection, RAG Knowledge Poisoning, and Multi-Step Agent Exploitation.

Start Challenges Scoreboard
Direct & Indirect Injection

Bypass system prompt guardrails, manipulate conversational context, and inject instructions through external documents.

RAG & Knowledge Poisoning

Exploit weak trust boundaries in vector databases, knowledge ranking, and unvalidated retrieval pipelines.

Agent Tool Misauthorization

Trigger unauthorized backend tools, IDORs, and chained multi-step agent escalation workflows.

Live DeepSeek LLM Backend

Interact with live AI models and deterministic challenge solvers in real-time containerized environments.