Your mandates are confidential by nature. TailorAI is built so that no data leaves a French perimeter : self-hosted open-weight models, inference on a private network, strict isolation in the database. An architecture that first reduces your exposure to extraterritorial law.
Open-weight models, served by you.
TailorAI runs on open-weight models served by vLLM on our own infrastructure. No calls to a closed API, no dependence on a provider that could shut off the tap or change its terms.
The weights run where you control them. No black box whose location and jurisdiction are out of your hands.
Open weights · vLLM · no third-party closed API in the inference loop.
OVHcloud France, inference on a non-exposed vRack.
Inference runs on OVHcloud servers in France. The compute machines are not exposed to the internet: they only talk to your instance, over a private vRack network.
No prompt data, no criteria, no target transits through a service outside this perimeter. The attack surface is reduced to what you orchestrate.
OVHcloud · France datacentres · private vRack · inference not exposed to the internet.
Row-level isolation, native audit log.
Each organisation is isolated at the row level in the database (RLS). Your mandates, your targets, your passes are visible only to you. Isolation is enforced in the database, not just in the application.
Every access and every action is recorded in a native audit log. You know who saw what, and when: the traceability a due diligence demands.
Multi-tenant RLS · native audit log · GDPR-compliant.
Built for the requirements of origination and sourcing teams.
Data processed and hosted in France, a record of processing activities, data-subject rights respected.
Strict multi-tenant enforced in the database (RLS). No cross-organisation leak possible by construction.
Native audit log of every access and action: the basis for a calm due diligence.
For Enterprise deployments, a single-tenant instance or dedicated hosting can be discussed against your security requirements.