Private AI for law firms

Private AI for law firms, on a model we run ourselves.

NeoLaw gives law firms private AI on its own language model, which runs on GPU servers we operate in Limassol. Contract review, chat with a contract and questions about your mailbox run there by default. Strict private mode keeps every AI step there.

  • Private model in Limassol
  • Data hosted in the EU
  • ISO/IEC 27001 and ISO 9001
Where each request goes
Where each kind of NeoLaw request runs in the standard setup, with the changes strict private mode makes marked as deletions and insertions.
RequestRuns on
Contract review, redlines and chat with a contract Private model by default only
Questions about your Outlook mailbox Private model, in every mode
Reading scanned pages (OCR) Private model by default only
Legal Agent research questions Cloud model on Auto, or the model the lawyer picks Private model
Case-law search Cloud model for search and summaries Private model and private index
Timesheets, pre-bills and drafting interviews Cloud model Private model
Web lookups for timesheets Cloud provider’s web search Switched off
Private model · GPU servers we operate in Limassol Application, database and files · hosted in the EU

The problem

Why not just use ChatGPT for lawyers’ work?

Public chatbots are good at drafting and summarising, and many lawyers already use them. The question is where the client’s document goes. Paste a contract into a public chatbot and its provider processes it on the provider’s servers, under the provider’s terms, outside your firm’s control.

For a firm bound by professional secrecy, that is hard to justify. A general-purpose chatbot is also not built around Cypriot law, and it does not hand back your own .docx with Word tracked changes. NeoLaw is built for that work, and it reviews contracts on a language model we run ourselves.

Architecture

Confidential AI, by architecture.

Where a model runs decides who can read the text you give it. These are the parts of NeoLaw and where each one runs.

The private model.An open-weights language model on GPU servers we operate in Limassol. Contract review, chat with a contract and questions about your mailbox run on it by default.

Search and indexing.In strict private mode, an embedding model on the same servers indexes your documents and the Cypriot corpus. Files you attach to a chat are indexed there in every mode.

The application.The application, its database and your files are hosted in the EU. Our information security management system is certified to ISO/IEC 27001.

One setting for the whole deployment.An administrator sets the mode. Outside strict private mode, lawyers can choose a cloud model for general research. In strict private mode, the model picker offers only the private model.

No training on your documents.NeoLaw does not train or fine-tune models on your documents.

Strict private mode

Every step on the private model.

Strict private mode is an option for the whole deployment. An administrator switches it on, at onboarding or later. From then on:

  • Chat, analysis, OCR of scanned pages, search and indexing run on the private model
  • No request goes to a cloud AI provider
  • The model picker offers only the private model
  • Failed calls are not retried on a cloud model
  • Web search is switched off

If the private model is unavailable, NeoLaw tells you. It does not switch to a cloud AI service.

  1. Your document
  2. NeoLawApplication and data, hosted in the EU
  3. Private modelGPU servers we operate in Limassol Cloud AI providersClosed in strict private mode

An administrator sets the mode once for the whole deployment.

Access

Who can see what.

FAQ

Private AI for law firms FAQ

Is ChatGPT confidential enough for client documents?

That depends on the plan, the provider’s terms and your professional obligations. A public chatbot processes what you paste on its provider’s servers, under the provider’s terms, so many firms keep client documents out of it.

NeoLaw runs contract review, chat with a contract and questions about your mailbox on a private model we operate. Strict private mode keeps every AI step there.

Where is our data processed and stored?

The application, its database and your files are hosted in the EU. The private language model runs on GPU servers we operate in Limassol, Cyprus. Connections to the app pass through Cloudflare, which terminates TLS.

Outside strict private mode, some features send text to a cloud AI provider, for example Legal Agent questions on the Auto setting, timesheets, pre-bill review, drafting interviews and case-law search. In strict private mode, no request goes to a cloud AI provider.

Is our data used to train AI models?

No. NeoLaw does not train or fine-tune models on your documents.

Can our lawyers still use a cloud model when they want to?

Yes, outside strict private mode. A lawyer can choose a cloud model for a conversation, for example for general research. With strict private mode on, the model picker offers only the private model.

How is strict private mode switched on?

An administrator switches it on for the whole deployment, at onboarding or later. It then applies to every user in the deployment. To discuss it, write to info@neolaw.ai.

How does the private model compare with the largest cloud models?

It is smaller than the largest cloud models. NeoLaw splits contract work into clause-sized steps, so the model works on one part of the contract at a time. Outside strict private mode, lawyers can still choose a cloud model for open-ended research.

Does it work with Microsoft 365?

Yes. Redlines come back as Word documents with native tracked changes. Lawyers can also connect an Outlook mailbox once your Microsoft 365 administrator approves the connection. Questions that read the mailbox always run on the private model.

Which certifications does NeoLaw hold?

Our information security management system is certified to ISO/IEC 27001, and our quality management system to ISO 9001.

How do we get started?

Request access at chat.neolaw.ai/register. Our team activates new accounts after a manual approval. Strict private mode is switched on by an administrator for the whole deployment.

Run your next review on a private model.

Contract review, chat with a contract and questions about your mailbox run on our private model by default. Questions about privacy or strict private mode? Write to info@neolaw.ai.

Create an account and our team will activate it.