Skip to content

Research & field notes

Prompt injection, model vulnerabilities, and defense strategies for production AI systems — from the team building the ZeroLeaks agent.

AI Security8 min read

How We Test Agent Boundaries (And What Our Score Does Not Mean)

Our methodology for continuously red-teaming autonomous agents: the boundaries we test, how we run attacks without putting production at risk, how we verify fixes, and an honest account of what our risk score is and is not.

Read the whitepaper→
AI Security9 min read

Securing AI Systems Against Prompt Extraction

A technical whitepaper on how system prompts, tools, and retrieval leak under adversarial pressure — and how ZeroLeaks tests that perimeter the way a real attacker would, continuously and automatically.

Read more→

Ready to secure your
AI agents?

Start free. Your first scan runs in minutes.

Start Your Free Trial