Bedrock Guardrails with DeepSeek

Yes, you can use Amazon Bedrock Guardrails with the DeepSeek-R1 model. Amazon Bedrock Guardrails provide configurable safeguards to help safely build generative AI applications at scale, and these guardrails can be applied to DeepSeek-R1 model deployments.

For DeepSeek-R1 deployments on both Amazon Bedrock Marketplace and SageMaker JumpStart, Bedrock Guardrails are supported. However, it's important to note that at the time of the most recent information, for these specific deployments, Bedrock Guardrails supports only the ApplyGuardrail API.

You can create multiple guardrails tailored to different use cases and apply them to the DeepSeek-R1 model. This helps in improving user experiences and standardizing safety controls across your generative AI applications. The guardrails can be used to introduce safeguards, prevent harmful content, and evaluate the model against key safety criteria.

Some key guardrail policies that can be applied include content filters, topic filters, word filters, and sensitive information filters. These guardrails are part of a comprehensive set of security features offered by Amazon Bedrock, which also includes data encryption, access controls, and secure connectivity options.

When using DeepSeek-R1 with Bedrock's playground or the InvokeModel API, it's recommended to use DeepSeek's chat template for optimal results. Additionally, implementing these guardrails is considered a best practice, especially since DeepSeek-R1 is an emerging model, and robust safety protections are crucial for responsible AI deployment.
Sources
Protect your DeepSeek model deployments with Amazon Bedrock Guardrails | AWS Machine Learning Blog
DeepSeek-R1 model now available in Amazon Bedrock Marketplace and Amazon SageMaker JumpStart | AWS Machine Learning Blog
DeepSeek-R1 models now available on AWS | AWS News Blog