SECURITY · DEDICATED HOSTING AND SOVEREIGN DEPLOYMENT
Our technology, deployed on your infrastructure.
For organizations with extreme sovereignty requirements, defense, security services, operators of vital importance, listed groups with high exposure, we deploy our simulation capabilities directly on your infrastructure. Air-gapped isolation, local open source LLMs, complete openness in model choice.
THE COMMERCIAL SHIFT
Some organizations structurally cannot use a shared SaaS configuration.
Our sovereign SaaS offering (infrastructure hosted in France, European capital and governance, no extraterritoriality) meets the requirements of the vast majority of our large-account clients. Some organizations, however, have higher requirements: European defense works under security classifications that tolerate no remote access; intelligence services cannot expose their strategic questions to a third-party system, whatever its location; operators of vital importance are bound by exclusive-control obligations; listed groups preparing a major financial transaction refuse to let those subjects pass through a third party, even a sovereign one.
For these organizations, the question is no longer where our servers are, but whether we can deploy on theirs. We stop being the operator: we provide the technology, the client assumes exclusive control. This configuration mobilizes their teams, their infrastructure and their governance, but it delivers the level of sovereignty these organizations cannot obtain any other way.
THREE LEVELS OF SOVEREIGN DEPLOYMENT
Adapted to each organization's specific requirements.
Level 1: Sovereign SaaS.
Our default offering, described in detail on the Sovereignty and infrastructure page. Infrastructure hosted in France by qualified French providers, European capital and governance, no extraterritoriality. This configuration suits the vast majority of our large-account clients.
Level 2: Dedicated hosting on client infrastructure.
Our technology is deployed directly on the client's infrastructure: private data centers, controlled sovereign cloud, on-premises. The client keeps exclusive control of the entire stack, hosting, network, data, processing. Our teams support the deployment with circumscribed, logged access; remote connections can be limited to maintenance windows, or removed. Suited to operators of vital importance, high-exposure groups and sensitive public authorities.
Level 3: Air-gapped deployment with a local open source LLM.
Our technology is deployed on infrastructure physically isolated from any external network, including the internet. It relies exclusively on open source language models deployed locally: Mistral, LLaMA, Qwen, DeepSeek, or any other relevant model. No external connection is required during operation. Suited to the defense industry, security and intelligence services, and any organization requiring absolute sovereignty.
A STRUCTURAL COMMERCIAL DIFFERENTIATOR
We do not lock our clients into any particular language model.
Unlike most AI decision-intelligence platforms, Imagine All The People does not lock its clients into one or two proprietary language models. Our architecture is structurally LLM-agnostic: we run the language model the client chooses, in the deployment configuration that matches their requirements.
Proprietary models in sovereign SaaS configuration.
We run the market's leading proprietary models, OpenAI's GPT, Anthropic's Claude, Google's Gemini, according to the client's preferences, ensuring that orchestration stays under our direct control and that prompts and instructions remain in Europe as much as possible. Clients who prioritize the raw performance of the most advanced models can choose this configuration.
Sovereign European models.
We also run the emerging sovereign models: Mistral, LightOn, Kyutai and other sovereign AI players. This configuration particularly suits clients who want to strengthen the sovereign dimension of their use while keeping the performance of recognized models.
Open source models deployable locally.
In air-gapped configuration or according to the client's preferences, we run open source models deployed locally: Mistral in its open source version, Meta's LLaMA, Alibaba's Qwen in its open source version, DeepSeek, Microsoft's Phi, or any other relevant open source model. This configuration guarantees total independence from any third-party model provider and forms the foundation of the most demanding air-gapped deployments.
COMPARING THE THREE DEPLOYMENT LEVELS
Matching the deployment level to the sovereignty requirement.
| Sovereign SaaS | Dedicated hosting | Air-gapped with local LLM | |
|---|---|---|---|
| Location of the infrastructure | Hosted in France by qualified French providers | Client infrastructure, private data centers or controlled sovereign cloud | Client infrastructure physically isolated from any external network |
| Operator of the processing | Imagine All The People | The client, with support from our teams | The client, autonomously after initial deployment |
| Language models available | Proprietary (GPT, Claude, Gemini), European (Mistral, LightOn) | Proprietary, European, or open source per the client's preferences | Exclusively open source deployed locally (Mistral OSS, LLaMA, Qwen, DeepSeek) |
| External connections required | Standard connections to the model providers | Connections limited to maintenance per the client's requirements | None, total physical isolation |
| Typical target clients | Major brands, consultancies, standard listed groups | High-exposure listed groups, operators of vital importance | Defense, security services, classified authorities |
Does your organization require exclusive control of its sovereignty?
The executive leadership of a defense prime contractor, a security service, an operator of vital importance, the leadership of a major financial transaction: our team can scope with you the parameters of a deployment adapted to your specific requirements. From the initial qualification of your needs to operational activation, allow three to six months depending on the deployment's complexity and the required level of isolation.
See our sovereign SaaS offering →