An ecommerce firm is deploying a chatbot that answers product questions and returns order details. They need safeguards to filter harmful content from user inputs and from the bot's responses. Which Bedrock feature or resource should they use?

Amazon Bedrock Guardrails are specifically designed to implement safeguards for generative AI applications. They allow you to define policies to detect and filter harmful content in both user inputs and model responses, ensuring the chatbot adheres to safety guidelines. Amazon Bedrock Agents are used for orchestrating multi-step tasks and connecting to external systems, not primarily for content moderation. Amazon Bedrock inference APIs are the programmatic interfaces for interacting with models, but they don't inherently provide content filtering. Amazon Bedrock custom models refer to fine-tuned models, which can improve performance but don't automatically include safety guardrails.