Guardrails for Amazon Bedrock enables customers to implement safeguards in generative Al applications that are customized to their use cases and responsible Al policies. Guardrails supports multiple filtering policies that enable customers to better control and customize generative Al applications based on their safety and privacy requirements.
First, Guardrails provides configurable thresholds to filter content across six harmful categories - hate, insults, sexual, violence, misconduct (incl. criminal), and prompt attack (jailbreak and prompt injection). Second, customers can define a set of topics that need to be avoided in the context of their application (e.g., investment advice in a banking assistant). Third, it supports detection and removal of Personally Identifiable Information (PII) to protect users' privacy. Fourth, Customers can provide a set of words to be blocked in their generative Al applications (e.g., profanity, competitor names). Fifth, Guardrails also supports contextual grounding checks to detect and filter hallucinations in model responses for RAG (retrieval-augmented generation) and summarization use cases. Lastly and most importantly, Guardrails supports the ApplyGuardrail API to independently evaluate user inputs and model responses for any custom and third-party FMs (beyond the ones available in Amazon Bedrock). Customers can use the ApplyGuardrail API to evaluate prompts or responses for any FM, enabling centralized safety and governance for all your generative Al applications.
See this detailed video with a demo to learn how you can use Guardrails for Amazon Bedrock to implement a comprehensive set of safety and privacy controls to manage user interactions in generative Al applications.
Learn more about Guardrails for Amazon Bedrock: [ Ссылка ]
Subscribe:
More AWS videos: [ Ссылка ]
More AWS events videos: [ Ссылка ]
Do you have technical AWS questions?
Ask the community of experts on AWS re:Post: [ Ссылка ]
ABOUT AWS
Amazon Web Services (AWS) is the world’s most comprehensive and broadly adopted cloud platform, offering over 200 fully featured services from data centers globally. Millions of customers — including the fastest-growing startups, largest enterprises, and leading government agencies — are using AWS to lower costs, become more agile, and innovate faster.
#FoundationModels #Guardrails #Hallucinations #ContextualGrounding #Groundedness #DataGovernance #ResponsibleAI #AWS #AmazonWebServices #CloudComputing
