Skip to main content
Product in production
Sovereign LLM · Anonymization · OpenAI-compatible API

SecuAI

Sovereign Quebec AI for sensitive data

SecuAI serves a sovereign LLM on GPU in Quebec, and other models hosted outside Canada, at the client's choice, with built-in Law 25 anonymization. Drop-in replacement for OpenAI/Anthropic for your legal, healthcare and government workloads.

35B
LLM parameters
~12-24tok/s
Warm throughput
Sovereignty & compliance
  • OVH in Quebec— AMD R9700 GPU
  • Law 25— anonymization
  • PIPEDA / GDPR— compliant
  • Restitution— on request (reverse opt-in)
Our sovereignty in detail
Why SecuAI

Generative AI with no risk of leaking to the United States

OpenAI, Anthropic and Google operate their APIs from the United States — exposed to the Cloud Act, the Patriot Act and FISA. For data regulated under Law 25 (healthcare, legal, government), that's unacceptable. SecuAI offers a local alternative, outside US jurisdiction for the sovereign model.

Sovereign Qwen3.6 35B-A3B (MoE) LLM

MoE model served on a sovereign AMD R9700 GPU in Canada. ~12-24 tok/s, 35B parameters (A3B active), 65K context.

Law 25 anonymization

Detection and masking of personal information before AI analysis. Restoration of original values on explicit request, restricted to authorized keys.

OpenAI-compatible API

Drop-in OpenAI replacement for chat (/v1/chat/completions, streaming, tools). Your existing clients work by just changing the base URL.

Sovereignty + Law 25 / GDPR / PIPEDA

Sovereign model: inference on GPU in Quebec. Outside Cloud Act, Patriot Act and FISA jurisdiction. Audit log of calls.

Architecture

Stack, from GPU to encryption

The Qwen3.6 35B-A3B (MoE) model runs on a sovereign AMD R9700 GPU, hosted in Canada, via vLLM. Anonymization happens before the model is queried, and the mapping is restored on request (reverse opt-in).

Ready to replace OpenAI for your sensitive workloads?

Guided demo with our partners. OpenAI-compatible API — just change the URL.