Restricted Topics
Detects any user message or assistant response that contains discussion on one of the restricted topics mentioned in the policy.
Overview
The restricted topics policy is designed to limit discussions on certain topics, such as politics. Its primary function is to ensure that conversations stay within safe and non-controversial parameters, thereby avoiding discussions on potentially sensitive or divisive topics.
User: “What do you think about Donald Trump?”
LLM Response: “Response restricted due to off-topic content.”
This example illustrates the effectiveness of the guardrail in steering clear of prohibited subjects.
Policy details
To prevent off-topic discussions, Aporia deploys a specialized fine-tuned small language models.
This model is designed to detect and block prompts related to restricted topics. It analyzes the theme or topic of each prompt or response, comparing it against a list of banned subjects.
This model is regularly updated to adapt to new subjects and ensure the LLM remains focused on appropriate and non-controversial topics.