How to Secure Vertex AI Training Jobs with Regional CMEK?

PCSE By islever Editorial Last updated

Short answer

C Configure CMEK for code, training data, and exported models, as Vertex AI automatically encrypts operational metadata with Google-managed keys.

Question

You are working with developers to secure custom training jobs running on Vertex AI. For compliance reasons, all supported data types must be encrypted by key materials that reside in the Europe region and are controlled by your organization. The encryption activity must not impact the training operation in Vertex AI. What should you do?
  1. Encrypt the code, training data, and metadata with Google default encryption. Use customer-managed encryption keys (CMEK) for the trained models exported to Cloud Storage buckets.
  2. Encrypt the code, training data, metadata, and exported trained models with customer-managed encryption keys (CMEK).
  3. Encrypt the code, training data, and exported trained models with customer-managed encryption keys (CMEK). Correct Answer
  4. Encrypt the code, training data, and metadata with Google default encryption. Implement an organization policy that enforces a constraint to restrict the Cloud KMS location to the Europe region.

Key Takeaways

This question tests the precise scope of CMEK coverage in Vertex AI, where candidates often mistakenly assume metadata is also encrypted with customer-controlled keys.

Securing Vertex AI custom training jobs requires Customer-Managed Encryption Keys (CMEK) for data and models while acknowledging that operational metadata remains protected by Google-managed keys. The community consensus confirms option C aligns with Google's official documentation on CMEK scope.

Common mistake: Option B is frequently chosen because it attempts to apply CMEK to all components including metadata, but Vertex AI automatically encrypts operational metadata with Google-managed keys regardless of CMEK configuration.

Expert Analysis

Why the Answer Is Correct

Option C correctly applies Customer-Managed Encryption Keys (CMEK) to code, training data, and exported models, satisfying the compliance requirement for regional key control. Google Cloud explicitly states that CMEK covers these core assets without interfering with Vertex AI's underlying training operations. By utilizing a Cloud KMS key located in Europe, you maintain organizational control over encryption materials while preserving job performance. This approach directly addresses the prompt's constraints regarding data residency and operational impact.

Why the Other Options Are Wrong

Option A fails to cover all required assets, leaving code and training data unprotected by customer-controlled keys. Option B incorrectly assumes CMEK extends to operational metadata, which contradicts Google's architecture documentation. Option D relies on Google-managed encryption keys, which violates the explicit requirement for organization-controlled key materials. Additionally, restricting KMS locations via organization policy does not guarantee that Vertex AI will use those keys for the specified resources.

Community Comment Notes

Multiple community members highlighted the official documentation stating that CMEK does not encrypt operational metadata like job names or regions. Comments [3], [4], and [5] emphasize that metadata is consistently handled by Google-managed keys to maintain system stability. Users who selected option B often overlooked this architectural boundary, leading to incorrect assumptions about full encryption coverage. The consensus strongly validates option C based on verified product specifications.

Official Reference

https://cloud.google.com/vertex-ai/docs/general/cmek#resources

This link opens the official vendor documentation in a new tab.

Exam Strategy

Always verify the exact scope of managed services' encryption features before selecting options that claim comprehensive customer key coverage. For Google Cloud security exams, memorize which resources are explicitly excluded from CMEK or VPC Service Controls to avoid over-provisioning traps.

Frequently Asked Questions

Why isn't metadata encrypted with CMEK in Vertex AI?

Vertex AI automatically protects operational metadata using Google-managed keys for performance and consistency, so CMEK only covers user data and models.

Can I enforce the Europe region restriction for CMEK without impacting training jobs?

Yes, creating a regional Cloud KMS key ring in Europe and linking it to your Vertex AI project satisfies compliance without adding latency to training workloads.

More PCSE FAQ →

Community Votes

C
44%
B
33%
D
22%

44% of anonymous learners picked answer C. Votes are pick records left by other test-takers — they are not the verified answer.

Discussion 8 contributions

Compiled from public community threads and reader submissions, lightly edited for clarity. Handles and vote tallies are shown as recorded, and comments reflect the views of individual contributors.

From public community threads 8

  1. dat987 👍 2
    Answer is C The CMEK key doesn't encrypt metadata, like the instance's name and region, associated with your Vertex AI Workbench instance. Metadata associated with Vertex AI Workbench instances is always encrypted using Google's default encryption mechanism.
  2. 3fd692e 👍 2 Answered B
    B is correct. D looks good but uses Google Managed Encryption Keys which violates the requirement of control the encryption resources outlined in the question.
  3. BondleB 👍 2 Answered D
    Option D enforces that all supported data types must be encrypted by key materials that reside in the Europe region.
  4. kalbd2212 👍 1 Answered C
    Ans is C Guys before recommending an answer please read the doc. In general, the CMEK key does not encrypt metadata associated with your operation, like the job's name and region, or a dataset's display name. Metadata associated with operations is always encrypted using Google's default encryption mechanism. https://cloud.google.com/vertex-ai/docs/general/cmek#benefits
  5. Zek 👍 1 Answered C
    C sounds right https://cloud.google.com/vertex-ai/docs/general/cmek#resources In general, the CMEK key does not encrypt metadata associated with your operation, like the job's name and region, or a dataset's display name. Metadata associated with operations is always encrypted using Google's default encryption mechanism.
Show 3 more comments
  1. Pime13 👍 1 Answered C
    In general, the CMEK key does not encrypt metadata associated with your operation, like the job's name and region, or a dataset's display name. Metadata associated with operations is always encrypted using Google's default encryption mechanism. https://cloud.google.com/vertex-ai/docs/general/cmek
  2. nah99 👍 1 Answered C
    C seems best. NOT B: "In general, the CMEK key does not encrypt metadata associated with your operation" NOT D: "If you want to control your encryption keys, then you can use customer-managed encryption keys (CMEKs) " https://cloud.google.com/vertex-ai/docs/general/cmek#resources
  3. yokoyan 👍 1 Answered B
    I think it's B.

Reader comments & corrections

No reader corrections yet — spotted an error or have a note? Share it below.

Log in to comment, report an error, or add a note about this question.

How to Implement TLS Interception for On-Premise Traffic in GCP

S-grade

How to Alert on Suspicious Outbound Traffic Using Threat Intelligence?

S-grade

How to Securely Host Databases Without Direct Internet Access?

S-grade

How to Manage Corporate and Public User Identities in Google Cloud?

S-grade

← Back to PCSE Study Guide