Complete records
Scored case scenarios and unscored research notes remain distinct.
AI SECURITY REFERENCE LIBRARY
AI security use cases and incident reference library
Explore 64 public-evidence analyses as threat-specific reference collections. Every case links to its sources, deployment boundary, rubric evidence, uncertainty, and—where the evidence permits—an indicative AITBM risk scenario.
Searchable themes
Threat and architecture collections built from the reviewed library.
Evaluation rubrics
Evidence routes connect incidents to reusable safety and security tests.
How to use the library
Search by a named incident when you need a source-backed reference. Browse a topic when designing a test plan or threat model. Open the complete record before quoting a score: unknown evidence, system boundaries, and uncertainty determine what the scenario can support.
Browse by threat and system type
41 REFERENCES
Agentic AI security
Public-evidence agentic AI security cases covering autonomous actions, tool use, identity, memory, delegation, and containment failures.
Open topic collection →4 REFERENCES
MCP security
MCP security case studies covering malicious servers, unsafe tool configuration, command execution, credential theft, and agent hijacking.
Open topic collection →7 REFERENCES
Prompt injection
Indirect and direct prompt injection case studies across copilots, connectors, browser agents, assistants, RAG, and operating-system AI.
Open topic collection →18 REFERENCES
AI data exposure
AI data exposure cases involving connectors, cross-tenant access, credentials, local files, private prompts, cloud data, and exfiltration paths.
Open topic collection →8 REFERENCES
AI supply-chain security
AI supply-chain security cases covering malicious models, packages, extensions, routers, agent configuration, dependencies, and namespace reuse.
Open topic collection →10 REFERENCES
AI coding agent security
AI coding agent security cases involving repositories, IDE features, CI secrets, packages, shell execution, database deletion, and developer credentials.
Open topic collection →5 REFERENCES
RAG security
RAG security cases covering retrieval authorization, vector databases, poisoned content, cross-user data, prompt-layer compromise, and cloud takeover paths.
Open topic collection →10 REFERENCES
AI model security
AI model security cases covering malicious model loading, inference backdoors, tokenizer blind spots, exposed inference, namespace reuse, and distillation.
Open topic collection →From reference to evaluation
A case study identifies a plausible path and the evidence that exposed it. An assessment must retest that path against your own deployment, include architecture-specific controls, derive operational risk from the real dependency graph, and use current evidence confidence.