Does Hukum-AI work in Arabic?

Yes. It is Arabic-native: retrieval, chunking and answer generation are built for Arabic legal language, with English supported in the same index.

Where is the data stored?

Inside the customer's own environment. Hukum-AI is deployed in-tenant, so the document corpus and the audit log stay under the customer's control and residency rules.

Can it cite its sources?

Every answer carries inline citations that resolve to the exact passage used. If no supporting passage is retrieved, the system reports that it cannot answer.

Is it PDPL compliant?

It is designed to align with PDPL and NDMO data-management practice, and deployment keeps personal data inside the customer's environment. Formal compliance is assessed per deployment with the customer's DPO.

Does it replace lawyers?

No. It shortens research, not judgement. Output is drafted for a qualified reviewer.

A scoping call, then a proof of concept on a defined corpus and a real set of questions from your team.

Product

Hukum-AI

Arabic legal and regulatory research with citation-grounded answers

Hukum-AI is an Arabic-native legal and regulatory research system built by Siyada Tech. It answers questions over Saudi laws, regulations and internal policy documents, and returns every answer with links to the source passage it came from. It runs inside the customer's own environment with a full audit trail.

Last reviewed: 2026-08-08 — Siyada Tech engineering

What it is

A retrieval-augmented research assistant for Arabic and English legal and regulatory material. Users ask a question in natural Arabic or English; Hukum-AI retrieves the relevant passages from the connected corpus, drafts an answer, and attaches the citations that support each statement. Answers it cannot ground in a source are returned as unanswered rather than guessed.

The problem it solves

Saudi legal and regulatory material is Arabic-first, spread across issuing authorities, and revised often. General-purpose chat models handle Arabic legal language poorly, cannot see internal policy, and produce fluent answers with no verifiable source — which is unusable for legal work.

Who it is for

In-house legal teams at Saudi enterprises
Compliance and regulatory affairs functions
Government legal and policy departments
Law firms handling Saudi regulatory work

How it works

01Ingest: laws, regulations, circulars and internal policy are parsed, chunked and indexed with Arabic-aware segmentation.
02Retrieve: hybrid keyword and semantic retrieval over the Arabic and English index, tuned for legal terminology and diacritic variance.
03Ground: the model may only use retrieved passages; every claim is tied back to a passage identifier.
04Cite: the answer renders with inline citations that open the exact source text.
05Log: question, retrieved context, model version and answer are written to an audit log.

Deployment

In-tenant deployment inside the customer's cloud subscription or data centre
Private VPC and on-premises options where data cannot leave the environment
REST API and web interface; SSO via the customer's identity provider
Document sources connected read-only from SharePoint, network shares or a document management system

Security & compliance

Personal data handling designed to align with Saudi PDPL and NDMO data-management practice
Data residency stays wherever the tenant is deployed; no corpus leaves the environment
Role-based access control; retrieval respects the source system's permissions
Immutable audit log of prompts, retrieved sources, model version and outputs
Independent certification status: [EVIDENCE REQUIRED]

Evidence

We publish a metric only with its definition, method, sample size and evaluation date. Anything not yet measured to that standard is marked below rather than claimed.

Metric	Definition	Method	Sample size	Evaluated	Limitations
Citation accuracy on Saudi regulatory queries	[EVIDENCE REQUIRED]	[EVIDENCE REQUIRED]	[EVIDENCE REQUIRED]	[EVIDENCE REQUIRED]	[EVIDENCE REQUIRED]
Answer groundedness rate	[EVIDENCE REQUIRED]	[EVIDENCE REQUIRED]	[EVIDENCE REQUIRED]	[EVIDENCE REQUIRED]	[EVIDENCE REQUIRED]
Research time per matter	[EVIDENCE REQUIRED]	[EVIDENCE REQUIRED]	[EVIDENCE REQUIRED]	[EVIDENCE REQUIRED]	[EVIDENCE REQUIRED]

Limitations & what it does not do

Answer quality is bounded by the corpus supplied; Hukum-AI does not know about documents it has not been given.
It supports legal research; it is not legal advice and does not replace a qualified practitioner's judgement.
Very recent regulatory changes appear only after the corpus is refreshed.
Scanned documents require OCR quality sufficient for reliable Arabic extraction.
We do not claim zero hallucination. Grounding and citation reduce unsupported statements; measured rates are pending publication.

Frequently asked questions

See Hukum-AI on your own corpus

We run a scoped proof of concept against your documents and your real questions, then show the citation trail.