Agentic AI Security Training
AI Security Labs Platform
Hands-on labs for red-teaming autonomous AI systems. Learn to find the vulnerabilities automated scanners miss.
UPCOMING AI SECURITY LABS
Lab Catalog
Each lab simulates a real AI security engagement. Start anywhere — no prescribed order.
Lab 04: Cross-Session Memory
Investigate cross-session memory leakage in an internal AI helpdesk where one employee's conversation context may surface in another employee's session.
View objective →Lab 01: Asset Knowledge Poisoning
Investigate how malicious documents injected into an AI assistant's knowledge base can poison responses and exfiltrate data.
View objective →Lab 02: Hallucination Privilege Escalation
Exploit model hallucinations to bypass authorization checks and escalate privileges in an identity-aware AI system.
View objective →Lab 03: Supply Chain Compromise via MCP
Investigate a malicious Model Context Protocol (MCP) plugin that poisons the AI's tool calls and exfiltrates data.
View objective →Lab 05: Breakfun Financial — Data Exfiltration
Exploit a financial AI assistant's function calling to exfiltrate beneficiary and transaction data across tenant boundaries.
View objective →Lab 06: Marketing Relay — Agentic Orchestration
Investigate a multi-agent marketing workflow where a compromised sub-agent hijacks the orchestration flow.
View objective →
Why These Labs
Built by Practitioners
Designed by engineers who've run identity and data-governance systems at scale — and who compete in CTFs against live targets.
6
Planned Labs
1
Deployed
5
Upcoming