EXP.001EXPERIMENT
RAG SECURITYIN PROGRESS
RAG Poisoning
How vulnerable are RAG systems to malicious documents?
Research in progress · Results not yet availableRead the research brief
RESEARCH / OPEN QUESTIONS
Understand the failure modes.
Practical inquiry into the security of AI-native systems. We document the question, the approach, and the limits of what we know.
EXPERIMENTS
EVIDENCE BEFORE CLAIMSQuestions beyond the first experiment.
Areas we want to explore. These are research interests, not published studies.
01
Prompt Injection
AREA OF INTEREST02
Indirect Prompt Injection
AREA OF INTEREST03
Agent Tool Abuse
AREA OF INTEREST04
AI-generated Code Security
AREA OF INTEREST05
System Prompt Security
AREA OF INTEREST06