Dike
THE DIKE MANIFESTO

Grounding legal AI in authentic Arabic sources.

We're building the verifiable legal data infrastructure for the Arab world, connecting machine intelligence directly to primary statutory, regulatory, and judicial authority.

01. THE CORE THESIS

The translation gap & hallucination crisis

AI agents are already answering regulatory, corporate, and compliance questions across the Middle East. Yet almost every major model operates in an information vacuum: global legal APIs are built US-first, passing Arabic law through lossy English machine translation and discarding citation anchors.

Web search fares worse, returning outdated law firm blogs, commentary on repealed acts, and fragmented summaries with no statutory hierarchy. When an AI agent can't reach the official Arabic gazette or statute directly, it doesn't just stop. It hallucinates a plausible-sounding provision instead.

In legal applications, an invented statute or mistranslated article is not a minor imperfection; it is unacceptable risk. Dike solves this at the data layer.

02. FOUNDATIONAL PILLARS

How Dike approaches legal truth

Pillar 01

Verifiable Primary Authority over Synthetic Guesses

AI models generate plausible language; courts and legal practitioners require verifiable statutory truth. Every response generated with Dike anchors directly to an official source ID, gazette publication date, and authentic statutory text in its original Arabic.

Pillar 02

Arabic-Native, MENA-First Architecture

General-purpose legal APIs treat Arabic as a translation target. Dike is built natively for Arabic legal morphology, statutory hierarchy, and the distinctive judicial conventions of Saudi Arabia, the UAE, and Egypt, expanding across the GCC.

Pillar 03

Topological & Temporal Graph Integrity

Law isn't static text. It's an evolving graph of amendments, repeals, and cross-statutory citations spanning decades, and we model that full structure so AI agents never cite a repealed article or an outdated decree.

Pillar 04

Deterministic Grounding for AI Developers

We provide sub-millisecond REST endpoints, vector search indices, and structured citation anchors designed for seamless integration into RAG pipelines, autonomous agent tool calls, and LLM reasoning workflows.

03. JURISDICTIONAL SCOPE

Deep regional coverage at launch

Dike indexes primary legal authority across major MENA jurisdictions with full bilingual metadata and citation resolution:

SAPrimary Source
Saudi Arabia

Royal Decrees, Council of Ministers Decisions, Ministerial Regulations, Umm Al-Qura Official Gazette.

AEPrimary Source
United Arab Emirates

Federal Decrees, Cabinet Resolutions, Dubai & Abu Dhabi Local Legislation, Federal Supreme Court principles.

EGPrimary Source
Egypt

Primary Civil, Commercial & Penal Codes, Al-Jarida Al-Rasmiyya Official Gazette, Cassation & Supreme Court rulings.

04. EMPIRICAL RIGOR

Built on verifiable legal research

Our engineering is guided by empirical benchmarks across 1.44M+ legal documents. We extract topological citation graphs, measure statutory codification velocities, and publish open datasets on Hugging Face so the research community can verify and build upon our findings.

Measured Impact
84.2% → 0.6%
Reduction in legal hallucination rate when grounding via Dike API
Explore Research
05. EXPLORE THE PLATFORM

Start building with Dike

Dike: The Arabic Legal Source Layer for AI
Privacy Policy© 2026 Dike. All rights reserved.