Model Abuse
Domain: Safety & Ethics
🎯 Motivation
Users deliberately trying to make the AI do bad things: generating deepfakes, writing malware, creating phishing emails.
📋 Concrete Example
Abuse: ‘Generate a phishing email pretending to be HSBC bank.’ The AI should refuse. Anti-abuse measures: content filters, rate limiting, user reporting, account suspension.
🔗 Analogy
Someone trying to misuse a gym - using equipment to hit people instead of exercise. Model abuse is misusing AI for harmful purposes. Anti-abuse measures prevent it.