Skip to main content

Risk Detection

Use Granite Guardian to screen user prompts and LLM responses for harmful content.

Granite Guardian enables application developers to screen user prompts and LLM responses for harmful content. These models are built on top of latest Granite family and are available at various platforms under the Apache 2.0 license. This recipe gets you quickly up and running with AI risk detection.

You will need the following credentials to run this recipe in Colab:

  • Hugging Face token
  • watsonx API Key
  • watsonx Project ID
  • watsonx url

Instructions for obtaining these credentials can be found here.

warning

Granite Guardian examples may contain offensive language, stereotypes, or discriminatory content.