QuestionQ130

VerifiedAI Safety, Security, and Governance

A company is developing a GenAI solution to help children aged 8 through 14 learn about world history. The system must support concurrent use by up to 30 children in a classroom and return responses in under 1 second. The company uses Amazon Bedrock and Amazon Nova Pro with a knowledge base containing history textbooks written for students aged 6 through 21. An Amazon Bedrock flow ensures that the solution uses the knowledge base to generate each response. The company configures the prompt-response temperature to 0.7.

Early testing produces some responses that are historically inaccurate, violent, and unsuitable for children. A GenAI developer must prevent the solution from responding to prompts inaccurately or inappropriately.

Which solution meets this requirement?

Explanation

Amazon Bedrock Guardrails content filters can evaluate and block harmful model responses, including violent content. Configuring a high-strength filter with the block action and attaching it to the flow prevents unsuitable generated output from being returned. Lowering inference temperature produces more deterministic output by favoring higher-probability tokens, reducing random and inconsistent responses.

Learn more

Community Discussion

No comments yet. Be the first to start the discussion!