Question 1

Is SecuAI already available?

Accepted Answer

Yes. SecuAI has been in open beta since April 2026. GA expected Q3 2026, with an active base of Quebec MSPs and enterprises.

Question 2

Which models are available?

Accepted Answer

Qwen3 35B-A3B (flagship sovereign model) and Mistral Small 24B, multilingual embeddings, FR/EN specialized models. Evolving catalog through managed updates.

Question 3

Does my data leave Canada?

Accepted Answer

No. GPU inference, storage and logs are 100% at OVH Beauharnois — outside Cloud Act/Patriot Act/FISA jurisdiction. No third-party calls to OpenAI/Anthropic/Google.

Question 4

Is the API OpenAI-compatible?

Accepted Answer

Yes, drop-in replacement: just change the base_url. Compatible with OpenAI SDK v1+, LangChain, LlamaIndex, Haystack, Semantic Kernel.

Question 5

How does PII anonymization work?

Accepted Answer

Automatic detection before inference: names, emails, Canadian SIN, RAMQ HCN, IBAN, credit cards, addresses. Reversible HSM tokenization, strict/permissive modes.

Question 6

What is BYOE?

Accepted Answer

Bring Your Own Encryption: payload encrypted client-side before transit, decrypted only inside the OVH HSM. Never plaintext on SecuAI side. Immutable audit.

Question 7

Is there an MSP tier?

Accepted Answer

Yes. Native multi-tenant console, per-tenant isolation, quotas (req/min, tokens/day), white-label branding, distributor discount.

Question 8

Are my prompts stored?

Accepted Answer

By default no. Explicit per-tenant opt-in for debug/audit, configurable 24h-30d TTL. No model training on your data, ever.

Question 9

How does SecuAI fit into the suite?

Accepted Answer

AI backbone of SecuScan (findings analysis), SecuSIEM (incident summaries, remediation suggestions), SecuMon (forecasting anomalies). Open API for your own workloads.

Question 10

What is typical latency?

Accepted Answer

Chat inference: p95 < 800 ms time-to-first-token, ~12-24 tok/s sovereign output throughput. Embeddings: p95 < 150 ms. SLA available at GA.

Question 11

How does SecuAI feed SecuSIEM?

Accepted Answer

At SecuSIEM GA, SecuAI provides French-language incident summaries, remediation suggestions, SOAR playbook generation. Sovereign model, no data ever feeds external training.

Question 12

How do I get started?

Accepted Answer

Free 30-day trial via /solutions/secuai, or contact the team for a guided MSP demo and migration from a US provider.

Frequently asked questions about SecuAI