AI without exfiltration for your strategic assets.
Fully sovereign AI agent deployment for defence, energy, aerospace, sensitive industry. Local AI or qualified trusted cloud, no data to US providers, controlled encryption, NIS2 and national trusted cloud doctrine compatible.
For industrial players handling strategic data, exfiltration to a US cloud LLM is not an option. Colombani.ai deploys AI agents locally (Ollama, on-prem Mistral) on controlled infrastructure, with encryption at rest, complete audit trail and trusted cloud doctrine compatibility. Approach suited to OIVs and OSEs under NIS2, entities under classified contracts (IGI 1300) and sovereign industry actors.
Use cases
Patent monitoring and prior art
Automated analysis of patent databases (EPO, USPTO, WIPO) on your technology fields. Convergence detection, competitive mapping, alerts on new relevant filings. Local models to avoid exposing your research directions.
Large technical documentation analysis
Reference extraction, tables of contents, cross-mapping of specifications across thousands of pages (manuals, IEC standards, technical dossiers). Local model on your servers.
Assisted predictive maintenance
LLM analysis of operations logs and RETEX reports to spot failure patterns. Does not replace dedicated ML models; complements them in natural language.
Specifications and procedures drafting assistant
Assisted drafting of technical specifications, quality procedures, design dossiers. The writer stays responsible; AI structures and proposes. Local model for confidential or classified data.
Ethics and regulatory framework
NIS2 (OES, EES) — reinforced cybersecurity
Operators of Essential Services and equivalent entities face reinforced cybersecurity obligations and incident notification. Colombani.ai deploys AI agents compatible with this requirement: segmentation, monitoring, alert forwarding to your SOC.
Trusted cloud doctrine and SecNumCloud
For sensitive data, French doctrine requires hosting on a SecNumCloud-qualified provider (OVH, Outscale, Cloud Temple) or on-premise. Colombani.ai designs compliant architectures and documents the choice for ANSSI audit.
IGI 1300 and classified contracts
Companies under classified defence contracts face strict processing rules. No AI deployment without security officer validation and written framework. Colombani.ai can operate under this framework on unclassified perimeters or after proper clearance.
What we don't do
- Critical embedded system development: other expertise (avionics, flight control computers).
- Autonomous command or fire control functions: prohibited and undesirable.
- Handling classified assets without prior clearance and written framework.
- SecNumCloud audit of a hoster: qualifications issued by ANSSI and accredited evaluators.
How we work with you
Consulting: sovereign architecture
Local AI (Ollama, Mistral), trusted cloud, no exfiltration outside the perimeter.
Solutions: AI agents & security
Industry AI agent plugins, AI-assisted security audits.
Training: AI for technical teams
Qualiopi programs for engineering, R&D and industrial cybersecurity teams.
Frequently asked questions
Can you guarantee no data leaves my infrastructure? +
Yes, with a fully local architecture: model running on your GPUs (Ollama with Mistral, Qwen or Llama), no outbound connections, local logs, encryption of artefacts at rest. The trade-off is performance: local models are more limited than Claude or GPT-4 but sufficient for most industrial use cases.
How do you integrate AI with an air-gapped IS? +
The model is installed directly in the air-gapped IS with embedded weights. Updates happen through controlled transfer (validated physical media). No cloud dependency. Colombani.ai delivers everything as an installable package, tested on an isolated environment equivalent to yours.
Are you cleared for defence work? +
Colombani.ai currently operates on unclassified perimeters. For a contract under IGI 1300, a prior clearance procedure would be required. Contact us in advance to scope the terms.
Discuss your project
First call is free. Straight answer on what is feasible and what is not within your ethics framework.