Tools AI

Sovereign AI Placement Check

Where can each of your AI workloads safely run?

List the AI you use or plan, and the most sensitive data each one touches. See whether it belongs in a business AI product, a cloud AI service in the right region, a government cloud, dedicated GPUs or your own data center, and which ones break a rule today.

  • CUI & ITAR
  • HIPAA
  • GLBA
  • GDPR
  • EU AI Act
  • GPU sizing
  • 5 minutes
Map my AI How MALA helps

Step 1 · Your AI workloads

List what you run or plan, and what it touches

Start from the example closest to you, then change any row. For each workload, pick the most sensitive data it can see, including data people paste in.

Example

Workloads

Scale
Used to size GPUs for anything you would run yourself.
people
Larger open models answer better but need more GPUs.
How placement is decided
Business AI product ChatGPT Enterprise, Microsoft 365 Copilot, Claude for Work, Gemini for Workspace and similar, under business terms that bar training on your data
Cloud AI service Azure OpenAI, Amazon Bedrock or Google Vertex AI in a commercial region you choose, reached over private networking
Government cloud Azure Government or AWS GovCloud, for CUI and export-controlled data, with the model on the provider's authorized list
Dedicated GPUs GPU cloud or colocation reserved for you, for training, fine-tuning or open models you run yourself
Your data center Full control, for the most restricted data or when no provider meets the rule
GPU sizing For chat, search, coding and customer agents: roughly 1 GPU per 400 users for a small model, 2 per 250 for a medium model, 8 per 300 for a large one. Batch work such as analytics or document processing starts at one model's worth (1, 2 or 8 GPUs). Training or fine-tuning adds 8. Cost at $2.00 to $3.50 per GPU-hour, 730 hours a month

Step 2 · Your placement plan

Put every AI workload where your rules allow

Unlock the conditions each placement depends on, the questions to put to every AI provider before you sign, and a 90-day plan to move what is out of bounds.

Conditions for each placement

The contract terms and settings each workload depends on.

Questions for every AI provider

Where data is stored, who can see it, and whether it trains anything.

90-day plan

What to stop, move and stand up in each 30-day block.

No cost, no obligation. MALA is vendor-sponsored.

Next step

Put each AI workload where it belongs.

MALA prices government cloud regions, GPU cloud, colocation and private connections side by side, and brings in AI governance partners, so your AI runs where your rules allow. No consulting fee.

Request a 30-minute call Schedule a consultation

An advisor responds within one business day. Call +1 (603) 802-2469.