gws-modelarmor-sanitize-response

To sanitize AI model responses for outbound safety, you apply predefined Google Model Armor templates to filter user-facing content. This process removes or masks sensitive information and inappropriate language from the outputs before they reach end-users.

Response sanitization in AI safety is the process of filtering model outputs using predefined security and content policies. It ensures that AI-generated text is safe and appropriate for end-users by applying rules from a configured Model Armor template.

To apply a Model Armor template to filter chatbot content, you specify the full template resource name during the sanitization process. The template defines the security rules that automatically remove or mask inappropriate language from the response.

You need to use a template for content filtering when your application requires outbound safety for user-facing content. It is necessary when AI models might generate responses containing sensitive information or inappropriate language that must be masked.

Yes, response sanitization requires specifying a full template resource name. This pre-configured Model Armor template defines the exact security and content policies applied to filter sensitive or inappropriate information from the model outputs.

The best way to ensure outbound safety for generated text is template-based sanitization. By using pre-configured Model Armor templates, you consistently apply predefined security and content policies to filter user-facing content before it is sent.