使用 Amazon Bedrock Guardrails 构建安全可靠的生成式 AI 应用.pdf

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Z U R I C H|1 1 S E P T E M B E R 2 0 2 5 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Build safe and responsible Gen AI applications with Amazon Bedrock GuardrailsJordi PortaG A I 2 1 3Senior Solutions Archite

2、ctAWSSamuel RiedoMoe HaidarSolutions ArchitectAWSHead of Agentic AINexthink 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Introduction to Amazon BedrockThe need for AI guardrailsAmazon Bedrock GuardrailsFeature deep diveLearnings from NexthinkAgenda 2025,Amazon Web Services,Inc.

3、or its affiliates.All rights reserved.What is Amazon Bedrock?2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Amazon BedrockThe easiest and fastest way to build and scale generative AI applicationsApply safety,privacy,and responsible AI checksImplement safeguards customized to your

4、 application requirements and responsible AI policiesSecurely customize with your dataUse contextual information from your companys private data sources to deliver more relevant,accurate,and customized responsesOptimize for cost,latency,and accuracyScale cost-effectively with:Model distillationIntel

5、ligent prompt routingPrompt caching Build and orchestrate AgentsSeamlessly connect with company systems,APIs,and data sources to drive automationChoose the best modelHigh-performing FMs,including 1P Amazon models and 3P partner models Model choice features:Model EvaluationCustom Model ImportMarketpl

6、ace*Coming soonAWS security,privacy and reliability built-in 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.The need for AI guardrailsUndesired behaviors and outputs such as hallucinations are a major concern with generative AI models as they get increasingly creative and versati