Safety Considerations for Tool-Using AI
InterviewSystem Designtool useagentssafety
Related module: Alignment fundamentals
Problem statement
AI systems increasingly use external tools (web browsing, code execution, APIs).
- What new safety risks does tool use introduce?
- Design a permission system for an AI agent that can use tools.
- How should the AI handle tool use that might cause irreversible actions?