An Arabic-native legal-AI platform for Gulf law firms: autonomous agents, precedent mining, and handwritten-Arabic intelligence, built for data sovereignty.
Most legal AI is built for English-language, common-law workflows, then translated. Gulf practice is the inverse: Arabic-first documentation, multi-jurisdiction civil and Sharia-influenced law, and a deep reluctance to put client data on someone else's cloud. GulfLaw.ai starts from that reality.
It pairs an Arabic-tuned model with a retrieval layer over GCC case law and a suite of autonomous agents that always pass through a human gate, so the system drafts, researches, and monitors compliance, while a lawyer stays accountable for every output. Handwritten case notes (still 40% of Gulf practice) are photographed, read, and folded into a searchable knowledge base.
The whole platform is designed to run where the data must live: public cloud, private GCC data centres, hybrid, or fully sovereign on firm-owned hardware.
Turns a client inquiry into a structured case brief with conflict check and jurisdiction analysis.
Gate · partner approvalQueries precedent and live court databases, surfacing the most relevant cases with citation-chain mapping.
Gate · lawyer validatesGenerates contracts and motions from clause libraries, learning the firm's house style over time.
Gate · associate reviewReviews documents for relevance, privilege, and responsiveness at a fraction of manual cost.
Gate · QA samplingTracks regulatory change across five jurisdictions and flags active cases the change touches.
Gate · officer confirmsCaptures time from activity with transparent, tamper-evident logs to pre-empt disputes.
Gate · partner approvesManaged multi-tenant SaaS in-region, with SOC 2 / ISO 27001 controls.
Dedicated GCC data-centre deployment with stronger isolation and NIST CSF alignment.
Sensitive data stays on-prem; an air-gapped option keeps the most sensitive matters offline.
Fully firm-owned hardware with custom certification: zero external dependency.
GulfLaw.ai is an active experiment, not a shipped product. The hypotheses I'm probing: how accurately current OCR reads real Gulf-dialect handwriting; whether RAG over Arabic case law is faithful enough to trust with citation; how light a human-in-the-loop gate can be before accountability erodes; and whether a genuinely sovereign deployment can stay maintainable.
If you run a Gulf firm (or regulate them) and want to pressure-test these ideas, I'd like to talk.
Sovereignty, speed, and an Arabic-first system of record.
An exploratory prototype by Dr. Nabeel A. Khan. Figures are design targets, not audited results.