Projects
Open-source work around AI and agent security – standards, tools and deliberately vulnerable labs. More on GitHub.
OWASP AISVS
A structured, testable checklist for verifying the security of AI-driven applications – covering models, data, agents and MCP. Rico co-leads the project with Jim Manico and others.
Agentic Chaos
Inject reproducible faults – outages, latency, prompt injection, tool and memory poisoning, failing guardrails – and verify that your agent fails closed and your security invariants still hold.
damn-vulnerable-adversarial-agent
A Juice-Shop-style lab: a deliberately vulnerable bank assistant that demonstrates classifier evasion and fleet-wide blast-radius attacks against LLM agents with shared memory. Built for AI Agent Summit London.
AI Security Lab
A hands-on lab for AI security: adversarial evasion (PGD), robust training notebooks, a red-teaming mindset guide and a one-shot environment setup script.
OWASP AI Security Visualizer
An interactive force-directed D3.js graph of the OWASP AI and ML security landscape – guides, standards, tools and how they relate. Static, no CDN, community-curated data.