AWS has introduced cross-region inference (CRIS) for European customers, allowing secure access to generative AI models across multiple AWS Regions while complying with data protection regulations. The feature enables customers to route inference requests within predefined geographic boundaries, helping them build resilient applications that align with local data processing needs. This capability is designed to support businesses operating in the European Union by ensuring compliance with the General Data Protection Regulation (GDPR) and other relevant data protection laws. Source: awsml
CRIS is a managed capability in Amazon Bedrock that routes model inference requests within supported AWS Regions. Inference profiles define the Regions where requests can be routed, with system-defined profiles named after the model and the geographic Regions they support. These profiles help customers leverage AWS’s global-scale infrastructure to build generative AI solutions while maintaining data security and compliance. Global inference profiles route requests to any supported AWS commercial Regions, optimizing model throughput with minimal latency overhead. Requests are encrypted and remain within the secure AWS network, ensuring data remains protected during cross-Region operations. Source: awsml
The EU-specific inference profiles (EU CRIS) are designed to help customers navigate regulatory obligations by restricting traffic to European Union Regions. Requests from EU source Regions can only be routed to other AWS Regions within the European Union, while non-EU source Regions can still use EU CRIS for optimized inference within EU Regions. AWS encrypts data in transit between Regions and ensures all traffic remains on the AWS network, avoiding public internet exposure. Customers must explicitly specify a CRIS profile ID in their code when invoking models for cross-Region inference, as demonstrated in the provided code snippet. Source: awsml
