Guardrails
Rules and systems that constrain model inputs, outputs and actions.
In plain language
Guardrails include content filters, sensitive-data detection, permission control, tool allowlists, output validation and human review. They are not absolute safety but reduce obvious risk. For high-risk scenarios, default to refusal rather than relying on model compliance.
Example
A support agent must escalate to a human for refunds above 500 yuan.
Recommended tools
Related terms
Model Evaluation
Designing test sets and scoring rules to decide whether a model works for a specific task.
Reinforcement Learning from Human Feedback
Train a reward model from human preferences, then optimize model outputs with reinforcement learning.
Tool Use
A model using external tools such as search, code execution or APIs during reasoning.
Agentic Workflow
Breaking a task into steps planned by AI, executed by tools and reviewed by humans.