LLM Risks & Vulnerabilities

LLM Risk	Vulnerabilities	Description
Responsible AI Risks	Bias, Toxicity	Ensuring ethical model behavior by preventing discriminatory outputs and offensive content generation that could harm users or specific demographic groups
Illegal Activities Risks	IllegalActivity, GraphicContent	Preventing content that violates laws, promotes criminal behavior, or generates instructions for harmful activities that could endanger public safety
Brand Image Risks	ExcessiveAgency, Robustness	Protecting organizational reputation by avoiding misinformation, misattribution, and content that contradicts company values
Data Privacy Risks	PIILeakage, PromptLeakage	Safeguarding sensitive information by preventing the exposure of personal identifiable information and confidential data
Unauthorized Access Risks	UnauthorizedAccess	Securing systems by preventing exploitation of LLMs to gain unauthorized system access or execute malicious commands