Community Guidelines

Character.AI · Official agreement

Read the agreement

1 of 26 cards in Use & Processing

Use & Processing · Company practice

Illegal or harmful content is filtered by automated moderation and human review; AI models are designed with filters

The guidelines say Character.AI's systems filter illegal or harmful content through both automated moderation and human review. They also say its AI models are designed with filters and limits to prevent inappropriate outputs. The wording "designed with" describes the intent of the design; the page does not promise that no inappropriate output occurs.

Supporting quotations

Our systems filter illegal or harmful content through both automated moderation and human review
our AI models are designed with filters and limits to prevent inappropriate outputs

Qualifications & exceptions

  • Filtering is described for illegal or harmful content, using both automated moderation and human review.
  • Model filters and limits are described as "designed" to prevent inappropriate outputs.

Not stated here

  • The page does not say what content human reviewers see, when human review happens, or how long reviewed content is kept.
  • The page does not define "inappropriate outputs."
Also annotated in the same passage (2)