A financial firm requires its generative AI chatbot to produce accurate, verifiable answers for regulatory compliance. Which solution helps prevent the underlying foundation model from producing hallucinated (factually incorrect) responses?

Amazon Bedrock Guardrails are specifically designed to enforce policies and prevent generative AI models from producing undesirable content, including hallucinated or factually incorrect responses. By defining safety policies and content filters, Guardrails evaluate both user inputs and model outputs in real-time, ensuring that the chatbot adheres to regulatory compliance requirements and provides accurate, verifiable information. AWS Config is for auditing and evaluating the configurations of your AWS resources, not for controlling AI model output. Amazon Fraud Detector identifies fraudulent online activities, which is unrelated to AI model hallucination. AWS Audit Manager helps automate audit evidence collection for compliance, but it doesn't directly prevent AI model errors.