Generative AI Meets AWS Security

This article discusses the security considerations and implementations for Amazon CodeWhisperer Customizations, a feature that allows customers to customize CodeWhisperer using their private code repositories to receive more tailored and relevant code recommendations.

Specifically, the article covers:

• Design principles for CodeWhisperer Customizations, such as data isolation, single-tenant compute, and preventing unauthorized access
• Security controls implemented during customization management, including data encryption, compute isolation, and mitigating confused deputy risks
• Security controls during customization usage, such as authentication, authorization, compute isolation, and further confused deputy mitigations
• Conclusion highlighting the security invariants baked into CodeWhisperer Customizations from the start