Safety

AI guardrails

AI guardrails are controls that constrain, monitor, or validate model behavior before outputs or tool actions reach users or systems.

Expanded definition

Guardrails can include input filters, output validators, policy checks, retrieval constraints, tool permissioning, human approval, data loss prevention, and monitoring. They do not replace model quality or evaluation, but they reduce risk in production systems where generated output can affect users, data, or external tools.

Related terms

Explore adjacent ideas in the knowledge graph.

prompt injection safety policy tool use

Comparisons, tools, and models that connect to this idea.

Azure Openai Vs Amazon Bedrock (comparisons)
Generative Model (glossary)
Claude 3 5 Sonnet (models)
Adversarial Training (glossary)
Generative Adversarial Network Gan (glossary)
Graph Machine Learning (glossary)