Sovereign AI
Your data. Your models. Your boundary.
You decide where AI runs and what it may see. GridTailor enforces it for every integration, based on how the data is classified.
Where AI may run, by data classification
The platform refuses to send data to a provider its classification doesn't allow.
| Classification | Allowed AI providers | Must stay in your boundary | Audit kept |
|---|---|---|---|
| Public | Anthropic API, OpenAI API, Azure OpenAI, Amazon Bedrock (in boundary), Your private model | No | 1 years |
| Internal | Anthropic API, OpenAI API, Azure OpenAI, Amazon Bedrock (in boundary), Your private model | No | 1 years |
| Confidential | Anthropic API, Azure OpenAI, Amazon Bedrock (in boundary), Your private model | No | 3 years |
| CUI | Amazon Bedrock (in boundary), Your private model | Yes | 6 years |
Deployment options
In development
GridTailor-hosted, in boundary
Models on Amazon Bedrock inside GridTailor's AWS environment, using FIPS endpoints.
Planned
Your AWS account or GovCloud
Bedrock runs in your own account. GridTailor calls it through a role you control and can revoke.
Planned
Your own model
Point GridTailor at an open-weight model you host in your VPC or data center, reached through the Tailor Agent.
Later
Disconnected
The full stack, models included, deployed where there is no internet connection.
Our commitments
- Your prompts and data are never used to train models.
- Every AI request is recorded in your tamper-evident audit log.
- AI tools are read-only. Changes need a person's approval.
- No lock-in: model providers are swappable.
Handling CUI or other sensitive data?
Let's talk through where AI can and can't run for you.