Filtering unsafe input or output before it reaches users.
1 min readupdated 2026-07-04
/ quick answer
Moderation classifies text (or images) against categories like hate, violence, sexual, self-harm. Use it on both user input (block prompt injection) and model output (block leaks and hallucinations). Filtering unsafe input or output before it reaches users.
Filtering unsafe input or output before it reaches users. Moderation classifies text (or images) against categories like hate, violence, sexual, self-harm. Use it on both user input (block prompt injection) and model output (block leaks and hallucinations). In practice: OpenAI's moderation endpoint runs before every generation in a public chatbot. This dictionary node is part of the Onexial knowledge graph and links to related concepts, workflows and tools below.
Definition
Moderation classifies text (or images) against categories like hate, violence, sexual, self-harm. Use it on both user input (block prompt injection) and model output (block leaks and hallucinations).
Example
OpenAI's moderation endpoint runs before every generation in a public chatbot.
/ frequently asked
What is Moderation?
Moderation classifies text (or images) against categories like hate, violence, sexual, self-harm. Use it on both user input (block prompt injection) and model output (block leaks and hallucinations).
What is an example of Moderation?
OpenAI's moderation endpoint runs before every generation in a public chatbot.
Why does Moderation matter for AI and automation?
Filtering unsafe input or output before it reaches users. It connects to the workflows, prompts and tool stacks linked on this page, so you can move from definition to execution without leaving Onexial.