2.1 Given a scenario, use AI threat-modeling resources.

Resources

Open Web Application Security Project (OWASP) Top 10

OWASP publishes community top-10 risk lists that are widely used as checklists.

LLM Top 10

OWASP Top 10 for LLM applications: prompt injection, sensitive information disclosure, supply chain, data and model poisoning, improper output handling, excessive agency, system prompt leakage, vector and embedding weaknesses, misinformation, unbounded consumption (the 2025 list).

Machine Learning (ML) Security Top 10

OWASP list of risks to ML systems such as input manipulation, data poisoning, model inversion, membership inference, model theft, supply chain, transfer learning attacks, model skewing, output integrity and model poisoning.

Massachusetts Institute of Technology (MIT) AI Risk Repository

A free database that classifies AI risks from many frameworks by cause and domain (discrimination, privacy, misinformation, malicious use, safety). Useful for broad risk identification.

MITRE Adversarial Threat Landscape for Artificial-Intelligence Systems (ATLAS)

A knowledge base of real adversary tactics and techniques against AI, modelled on MITRE ATT&CK, with case studies. Use it to map attacks and plan detections.

Common Vulnerabilities and Exposures (CVE) AI Working Group

CVE program group working on how AI-related vulnerabilities are identified and recorded, so AI flaws get consistent CVE entries.

Threat-modeling frameworks

Structured methods such as STRIDE, PASTA, attack trees or data-flow diagrams, applied to AI components: data pipeline, model, prompts, tools and outputs.