Interactive AI Security Learning

Explore the through simple, interactive scenarios and build your understanding of AI security.

Interactive

LLM04

Supply Chain

Vulnerable or tampered third-party models, data, packages, or services introduce unsafe content and behavior into an AI system.

Interactive

LLM05

Data and Model Poisoning

Tampered training data, fine-tuning processes, or model components manipulate a language model into producing incorrect, biased, or exploitable behavior.

Interactive

LLM08

Hidden Context Exposure

A language model exposes internal rules, tool settings, or hidden data that attackers can use to bypass restrictions or abuse the system.

Interactive

LLM09

Vector and Embedding Weaknesses

Retrieval or semantic caching selects similar content without adequate source and access checks, exposing private data or introducing untrusted information.

Interactive

LLM10

Improper Output Handling

Systems process language model output as web content, commands, or file paths without appropriate validation or encoding, causing unintended actions and security vulnerabilities.