Tool use turns a language model into an actor. Each exposed tool is a permission grant; the safety question is not 'can the model do X' but 'can any input make the model do X'. Scope tools tightly and log every call.
GlossaryPart of the BotDoors AI Safety & Compliance Glossary · 37 terms.
Safety
Tool Use
Tool use lets an LLM call external functions — search, code, APIs — to extend what it can do.