01
Precedent and knowledge search
Search your own precedents, templates, opinions and research memos in plain language, with answers that link to the source document.
Private AI · Legal
Private AI for law firms and in-house legal teams
Client confidentiality is the foundation of legal work, and public AI tools put privileged documents on someone else’s servers. We supply on-premise AI hardware and our partners deploy open-weight LLMs on it in your office, so client files never leave your building.
The risk
Why legal data should not go to public AI tools
Privileged and confidential client documents uploaded to public AI tools
Unclear retention and training terms at third-party AI providers
Client data from the EU, DIFC or ADGM crossing borders without safeguards
Ethical walls between matters are hard to enforce in shared cloud tools
Use cases
What teams run on it
Open-weight LLMs deployed by our partners, working on your documents inside your network.
02
Contract review
Summarise contracts, extract key clauses and compare them against your playbook. Lawyers check every flag before advice goes to the client.
03
Due diligence
Process large data-room document sets to find change-of-control, termination and liability clauses, with the documents kept on your own servers.
04
Drafting support
Produce first drafts of letters, clauses and memos from your firm’s templates and style, for lawyers to review and finalise.
Compliance context
Rules that shape where your data can go
On-premise AI helps support compliance by keeping processing on hardware you control. It does not replace your own legal and compliance assessment.
UAE PDPL (Federal Decree-Law No. 45 of 2021)
The federal Personal Data Protection Law applies to personal data processing in the UAE outside the financial free zones and sectors governed by their own data laws. It includes requirements for transferring personal data across borders, and on-premise processing reduces the transfers you have to assess and document. Firms licensed in the DIFC or ADGM fall under those centres’ own data protection regimes instead.
DIFC Data Protection Law (DIFC Law No. 5 of 2020)
Firms in the Dubai International Financial Centre follow their own data protection law, supervised by the DIFC Commissioner of Data Protection. Transfers of personal data outside the DIFC need an adequate destination or appropriate safeguards, so on-premise AI keeps that analysis simpler.
ADGM Data Protection Regulations 2021
Entities in Abu Dhabi Global Market follow the ADGM Data Protection Regulations 2021. Its guidance covers impact assessments for high-risk processing, security of processing and safeguards for international transfers, all of which are easier to evidence when AI runs in-house.
EU GDPR, for EU clients and data subjects
The GDPR can apply to organisations outside the EU that offer goods or services to people in the EU. Transfers of their personal data outside the European Economic Area need an adequacy decision or safeguards, and fewer third-party processors means fewer transfers to justify.
This is general information, not legal advice. Rules change and depend on your licence, location and data. Ask your legal or compliance adviser which rules apply to your organisation.
Recommended starting point
Most legal teams start with the team bundle
Typically 5–25 users, running open-weight LLMs of 7B–14B parameters, quantized. Every bundle is quoted to your users, documents and site.
Most requested
Typically 5–25 users
Team bundle
A first private AI deployment for one team, practice group or pilot project.
- Typical users
- 5–25
- GPU class
- 1–2 professional GPUs
- Typical model size
- 7B–14B, quantized
- Storage
- Local NVMe
- Form factor
- Workstation or 2U–4U server
FAQ
Private AI for legal: questions
Can we keep ethical walls between matters?
Yes. Our partners set up separate document collections with role-based permissions tied to your single sign-on. The assistant only searches collections the user can access, so a lawyer on one matter cannot retrieve documents from a walled-off matter.
Is a Team bundle enough for a law firm?
For a practice group or a firm of up to about 25 regular users, a Team bundle running a quantized 7B–14B open-weight model is usually a good start. Larger firms, or those wanting several use cases at once, typically choose the Department bundle.
Are our documents used to train the model?
Not unless you decide to fine-tune on them. By default the model reads your documents at query time through retrieval, and nothing is sent outside your network. Any fine-tuning happens on your own hardware, under your control.
Does this help with GDPR for EU clients?
It can. Keeping processing in-house reduces the number of third-party processors and international transfers you need to justify. It does not answer every GDPR question, so your data protection lead should still review lawful basis, retention and client terms.
Is this legal advice?
No. This page is general information, not legal advice. Your own data protection and professional-conduct advisers decide what applies to your firm. Private AI helps support compliance; it does not replace your own assessment.
Request a quote
Request a private AI quote for legal
Share your users, use cases and site details. No obligation, and we reply within one business day.