AWS Updates Amazon Bedrock With Custom Model Imports and Enhanced Security
AWS marks the one-year anniversary of Amazon Bedrock by introducing custom model imports, updated Titan embeddings, and new security guardrails to help developers build generative AI applications.
Amazon Web Services (AWS) is marking the one-year anniversary of its Amazon Bedrock generative AI platform with a major update that introduces new capabilities for developers. The fully managed service provides easy access to foundation models from leading AI vendors like Anthropic, Meta, and Cohere, as well as Amazon's own Titan models. AWS views Bedrock as the central hub where organizations come to build and deploy generative AI applications.
A key new feature is Custom Model Import, which allows developers to bring their own proprietary models into the Bedrock environment. This enables organizations in highly regulated industries like fintech and healthcare to integrate models trained on Amazon SageMaker directly into Bedrock's serverless infrastructure. Users can now easily compare their custom models against other supported foundation models and deploy them across multiple workflows.
The update also includes an upgraded Amazon Titan embeddings model that creates smaller vectors to accelerate retrieval augmented generation (RAG) and vector database performance. Additionally, AWS is rolling out new guardrails to improve security and control over AI outputs. These enhancements come as AWS intensifies its competition with cloud rivals Google and Microsoft in the rapidly expanding generative AI market.