AI serving the public, sovereign and audit-defensible.
AI consulting and deployment for ministries, local authorities, agencies and public institutions. Compatible with trusted cloud doctrine (SecNumCloud, RGS), respecting public procurement rules. Available via BOAMP public tenders.
Colombani.ai supports administrations and public institutions in deploying AI agents compliant with trusted cloud doctrine (SecNumCloud, RGS) and the AI Act high-risk framework for sensitive uses (citizen service, individual decisions). Available via public tenders: active BOAMP monitoring, up-to-date permanent documents, scoped DCE responses. Typical engagements: studies, audits, Qualiopi training, agent development in a sovereign framework.
Use cases
Multilingual document translation and synthesis
FR/EN/ES/DE translation, synthesis of European or international reports, executive summaries for decision-makers. Local or trusted cloud model depending on sensitivity.
Regulatory and parliamentary monitoring
Automated tracking of official texts (Official Journal, Légifrance, ministry sites, Parliament), synthesis by topic, alerts on relevant amendments.
Citizen service assistant
Support for citizen request handling at the front-office: reformulation, orientation to correct forms, response preparation. Final decision retained by the public officer.
Large-scale report and proposal analysis
Structured analysis of call-for-projects files, consultation responses, impact studies. Key point extraction, issue mapping, stakeholder mapping.
Ethics and regulatory framework
SecNumCloud and trusted cloud doctrine
French PM 2021 circular and central cloud doctrine: sensitive state data hosted on SecNumCloud-qualified or equivalent solutions. Colombani.ai designs agent architectures compatible with this requirement.
RGS — General Security Framework
Public online services must comply with RGS. Encryption, authentication, logging. Our agent deployments integrated with a public online service include documentation expected by ANSSI.
AI Act — high-risk for sensitive public uses
The EU AI Act classifies some public uses as high-risk: credibility scoring, prioritization of public service access, education. Obligations: documentation, conformity assessment, human oversight, EU database registration.
What we don't do
- Automated individual administrative decisions: framed by CRPA art. L.311-3-1 (citizen information obligation) and AI Act art. 5 for certain cases.
- Replacing sworn public officers: the AI agent stays a tool, the decision stays human.
- Social credit or citizen scoring systems: prohibited by AI Act art. 5.
- SecNumCloud qualification of a solution: issued by ANSSI and accredited evaluators.
How we work with you
Consulting: sovereign architecture
Local AI or qualified trusted cloud, SecNumCloud / RGS compliant.
Consulting: AI Act compliance
Risk classification of your AI uses and expected documentation.
Training: AI for public officers
Qualiopi programs for ministry teams, local authorities and public operators.
Frequently asked questions
How do you respond to a public tender? +
Colombani.ai actively tracks BOAMP (French public tenders) and holds up-to-date permanent documents (DC1, DC2, tax and social attestations, professional liability, Qualiopi certification). We respond directly or as part of a consortium depending on the tender. Contact us in advance to scope a partnership on a targeted tender.
Is SecNumCloud qualification possible for deployed agents? +
SecNumCloud qualification applies to a hosting provider, not to an application agent. Colombani.ai deploys agents on already-qualified infrastructures (OVH SecNumCloud, Outscale, Cloud Temple) or locally on your infrastructure. The overall public service compliance is thus ensured.
Is training funding available for public officers? +
Yes. Colombani.ai is Qualiopi-certified (NDA 11757549975). Training programs are eligible for the training plans of the three public functions (State, territorial, hospital) according to your administration internal rules.
Discuss your project
First call is free. Straight answer on what is feasible and what is not within your ethics framework.