Language Model Security DB
A top-level knowledge base for recurring LLM and agent security patterns, integration failure modes, and field-ready research links.
This page connects the security database to the AI Security Hub, where practical review paths and Sentinel links live together.
Pattern tracking
Organize repeated issues across prompt injection, tool abuse, unsafe retrieval, identity confusion, and runtime escalation.
Field context
Anchor abstract model-security patterns to practical incident analysis, advisories, and deployment lessons.
Operational use
Give engineering, security, and governance teams one vocabulary for prioritizing model and agent risks.
Review paths
Review model files, RAG, MCP, and agent security resources in one place.
Open the primary resource hub for the security database.
Review red-team playbooks for prompt and agent abuse cases.
Connect retrieval risk and factuality evaluation to model-security review.
Read coordinated disclosures and technical writeups relevant to agent and integration risks.