Warden Industries
ENFR
Request a quote

Compliance packages.

The number of simultaneous users per package is confirmed during sizing. The package quotation builds on the findings of the AI Compliance Diagnostic.

Law 25Privacy impact assessmentIntellectual propertyContractual confidentialityllama.cppGemma 4Ministral 3gpt-ossLinuxFull-disk encryptionOutbound traffic refused

Compliance packages

One all-in package: hardware, installation and compliance.

Hardware, models and number of users are confirmed in writing before installation.

PKG-01

Compliance Essential

Deployment

24 GB workstation on site

An engineering team, a design office or a small site of 1 to 5 people.

  • Linux workstation with a 24 GB NVIDIA GPU, installed on site
  • Gemma 4 26B A4B and Ministral 3 14B
  • Private chat interface on your internal network
  • Privacy impact assessment (PIA), AI use policy, incident register
  • 2 hour staff training and first annual review

Optional monthly support available

Request this offer
PKG-02

Compliance Pro

Deployment

32 GB workstation on site

A technical department, a plant or a firm of 5 to 20 people.

  • Linux workstation with a 32 GB NVIDIA GPU, installed on site
  • Gemma 4 31B, plus Gemma 4 26B A4B for fast answers
  • Search across your technical documents and procedures (RAG)
  • Everything in Essential, plus a centralized audit log
  • Training per workstation and first annual review

Optional monthly support available

Request this offer
PKG-03

Compliance Enterprise

Deployment

96 GB server

A multi-site organization or several plants, 20 people or more.

  • Server with a 96 GB NVIDIA GPU, on site or in colocation in Montréal
  • Gemma 4 31B and 26B loaded together, or gpt-oss-120b
  • Multiple users, single sign-on and agents
  • Everything in Pro, plus a PIA per department
  • Automated decision procedures and continuity plan

Optional monthly support available

Request this offer

Included in every package

The compliance file is part of the system.

Every package includes the hardware and the documents your organization needs to frame its use of AI.

  • 01Compliance assessment and privacy impact assessment (PIA)
  • 02Hardware installed on site
  • 03Written policies
  • 04Logging and audit
  • 05Staff training
  • 06First annual review
See the packages

Technical stack

Allied-origin models, on hardware you control.

  • ModelsGemma 4 (Google, United States), Ministral 3 (Mistral AI, France) and gpt-oss (OpenAI, United States), under Apache 2.0 licences
  • ExcludedModels from China-based labs and any model derived from them, whatever the licence
  • Inference enginellama.cpp (llama-server, MIT licence). No PyTorch code runs in production
  • Operating systemLinux, with full-disk encryption
  • NetworkFirewall that refuses outbound traffic by default, verifiable with a network capture
  • LocationOn your premises, or in a colocation facility in Montréal for the Enterprise package

Our method

A clear method, step by step.

Hardware, models and documents are defined with you and confirmed in writing before installation.

Request a quote

01

Compliance diagnostic

An inventory of the AI tools already in use, a light privacy impact assessment and an action plan.

02

Design and sizing

Hardware, models and number of users defined with you, then confirmed in a written quotation.

03

Installation and lock-down

The system is installed on your premises, encrypted and closed to outbound traffic by default.

04

Compliance file and training

Written policies, incident register, audit log and training for your team.

05

Annual review and support

A yearly review of the system and documents, with optional monthly support.

Frequently asked questions

Your questions, our answers.

Does Warden Industries provide legal advice?
No. We provide technology, configuration and documentation tools. Have the compliance documents reviewed by your own legal counsel.
Where is our data processed?
On hardware installed on your premises, or in colocation in Montréal for the Enterprise package. The firewall refuses outbound traffic by default.
Which models do you use?
Gemma 4 (Google), Ministral 3 (Mistral AI) and gpt-oss (OpenAI), under Apache 2.0 licences. Models from China-based labs are excluded.
Who sells and invoices the system?
Warden Industries sells, installs and invoices the system. Every engagement starts with a written quotation and a contract.
Do our drawings and processes leave the site?
No. The models run on your hardware and the firewall refuses outbound traffic by default, which helps you meet the confidentiality agreements you have with clients and partners.
How many people can use it at the same time?
The number of simultaneous users is confirmed during sizing, based on the hardware chosen.

Not legal advice

Warden Industries provides technology, configuration and documentation tools. We do not provide legal advice. The compliance documents we prepare with you are working documents: have them reviewed by your own legal counsel, and refer to the guidance of the Commission d’accès à l’information.

Official references

Let’s talk about your AI use.

Describe your team, your sites, the data you handle and the AI tools already in use. We reply by email with next steps and a written quotation.

Request a quoteinfo@wardenindustries.com

Every engagement starts with a written quotation and a contract with Warden Industries, which sells, installs and invoices the system.