With Guardrails for Amazon Bedrock, customers can implement safeguards across foundation models based on their use cases and responsible AI policies, tailoring interactions and filtering undesirable or harmful content for a relevant and safe user experience in generative AI applications.

<div>
<p>This article announces Guardrails for Amazon Bedrock, a new feature in preview that allows customers to implement safeguards across foundation models (FMs) based on their use cases and responsible AI policies.</p>
<p>Specifically, the article covers:</p>
<ul>
<li>Customers can create multiple guardrails tailored to different use cases and apply them across multiple FMs.</li>
<li>Guardrails enable customers to define denied topics, configure thresholds, and filter harmful content across categories like hate, insults, sexual, and violence.</li>
<li>Guardrails evaluate user queries and FM responses against the defined restrictions, helping to prevent undesirable content.</li>
<li>Guardrails are supported for English text-based FMs, fine-tuned models, and Agents for Amazon Bedrock.</li>
<li>Guardrails for Amazon Bedrock is available in preview in the US East (N. Virginia) and US West (Oregon) Regions.</li>
</ul>
</div>


Safeguard generative AI applications with Guardrails for Amazon Bedrock (Preview)

Related articles