Review Mode set 2 Question 30

[Nouman]

Inconsistent Answers, Totally irrelevant to the Question.

None of the Answer is adequate.

A financial services organization is developing a generative AI chatbot using Amazon Bedrock to answer confidential customer questions. The solution must implement layered protections, including jailbreak detection mechanisms, to prevent advanced prompt injection attacks targeting the model. Security engineers also require detailed audit records whenever safety rules intervene, such as when model input or output is blocked or modified. In addition, the architecture must support multi-Region resilience so the AI assistant continues functioning if a Region becomes unavailable.

Which architecture satisfies these requirements?

1. Leverage SageMaker Script Mode to fine-tune an LLM on Amazon EC2 instances, enabling custom training scripts to optimize model performance with flexibility.

2. Use SageMaker Training Jobs to fine-tune an LLM deployed through a custom API endpoint, automating the model training process with scalable resources.

3. Utilize SageMaker Studio for fine-tuning an LLM deployed on Amazon EC2 instances, simplifying the training process with an interactive and intuitive environment.

4. Configure SageMaker Autopilot to fine-tune an LLM deployed via SageMaker JumpStart, streamlining model customization with automatic setup and minimal user intervention.