A sovereign AI platform for a leading law firm.
How a leading law firm got AI for its documents, on closed servers, with no legal data leaving the building.

The brief.
No legal data leaves the building. Not for training, not for inference, not for storage.
Documents that answer back.
The platform helps lawyers work with documents more effectively, and learn continuously from the firm’s own data.
It classifies and summarises documents automatically, extracts clauses, searches across historical matters and retrieves relevant precedents.
Why it was hard.
Confidentiality first. Client documents are among the most sensitive data there is. Public AI services were out of the question.
Sovereign by design. Storage and processing had to stay under the firm’s full control, with data residency enforced.
Built for lawyers. The tool had to fit legal workflows, not a generic chat window.
What lawyers can do now.
Learning from the firm’s own work.
Models are fine-tuned or adapted on the firm’s own documents. Answers and suggestions therefore reflect its style, its risk preferences and its institutional knowledge.
Lawyers’ corrections and newly added documents feed back into the system, under controlled conditions, so it keeps improving without ever leaving the firm’s perimeter.
Five phases, security first.
We took a phased approach that kept security, compliance and user adoption at the centre of the project.
- 01
Secure sovereign infrastructure
A local, closed-server environment with strong isolation, encryption at rest and in transit, role-based access control and full audit logging. All AI workloads run inside this perimeter.
Closed servers - 02
Secure data foundation
Controlled ingestion, cleaning and indexing of the firm’s existing document repositories and matter data. Sensitive information stays on-premises throughout processing and enrichment.
Data stays in place - 03
Private AI capabilities
Private large language models and custom NLP components for document understanding, clause analysis, summarisation and semantic search. The models run entirely inside the closed environment and can be adapted further on internal data.
Private LLMs - 04
Lawyer-facing application
A secure web portal where lawyers query documents, generate structured insights, compare clauses and retrieve prior work. It is designed for legal workflows, not generic chat.
Secure portal - 05
Continuous learning loop
Feedback mechanisms and controlled fine-tuning let the system improve from lawyers’ corrections and newly added documents, while staying fully sovereign.
Improves with use
Built with, and secured by.
- Infrastructure
- Local closed servers, in a private sovereign environment
- Container platform
- Red Hat OpenShift
- AI and machine learning
- Private LLMs and custom NLP models
- Application
- NestJS, TypeScript, React
- Security and compliance
- End-to-end encryption, role-based access, audit trails, data residency controls
The result.
Faster with documents. Backed by everything the firm already knows.
The result is a practical AI capability that helps lawyers work faster with documents and draw on the firm’s accumulated knowledge, while meeting the strictest requirements for confidentiality and data sovereignty.
What IT22 can do for your organisation.
If you are exploring AI for sensitive documents or regulated data, and you need full control over where information is stored and processed, IT22 can design, build and operate a sovereign solution tailored to your environment and workflows.
- Custom software development
- Sovereign and on-premises AI platforms
- Managed private cloud and OpenShift
- Secure document and knowledge systems
