AWS Certified AI PractitionerFoundation ModelsMedium
A large enterprise is considering using a foundation model for various internal tasks, including document summarization, internal knowledge base Q&A, and code generation for developers. The security team has strict data residency and privacy requirements, stating that all sensitive company data must remain within the company's private cloud infrastructure and cannot be exposed to external services. Which deployment strategy is MOST appropriate for this scenario?
- ADeploying an open-source foundation model on-premises within the private cloud.
- BSending anonymized data to a third-party foundation model provider.
- CUtilizing a public cloud-based foundation model API with data encryption.
- DImplementing a hybrid cloud solution with sensitive data processed externally.
Show answer & explanationAnswer & explanation
Correct answer: A. Deploying an open-source foundation model on-premises within the private cloud.
The core constraint is that 'all sensitive company data must remain within the company's private cloud infrastructure and cannot be exposed to external services'. Deploying an open-source foundation model on-premises directly addresses this by ensuring complete control over data residency and processing within the private cloud, satisfying strict security and privacy requirements.
Why the other options are wrong
- B. Anonymization might not always be sufficient for 'sensitive company data', and sending to a 'third-party provider' still exposes data externally.
- C. Public cloud APIs, even with encryption, involve data leaving the private infrastructure, violating the 'cannot be exposed to external services' rule.
- D. A hybrid solution with 'sensitive data processed externally' directly contradicts the requirement that all sensitive data must remain within the private cloud.
On-premises Foundation Model Deployment
Running a foundation model entirely within an organization's own private data centers or private cloud infrastructure, without reliance on external cloud providers for model inference or data processing.
- Provides maximum control over data security, privacy, and sovereignty.
- Requires significant internal infrastructure and expertise.
- Can be costly due to hardware and operational overhead.
Memory trick: On-premises: Your castle, your rules, your data.