← Back to HAQQ Blog

Arabic Is Hard for AI. Legal Arabic Is Harder. Who's Actually Solving It?

By HAQQ Team · · 10 min read · Mena

A real Arabic legal-AI scene emerged in 2025–26 — Adel, Shwra, Mohamy, Arabic.ai. But run them through a four-corner test and nobody hits all four.

Why Arabic is genuinely hard for AI

Most 'AI lawyer' apps are English models in a trench coat. That works until the language fights back, and Arabic fights back hard.

It's written right-to-left, which breaks naive text pipelines. It's diglossic: the Modern Standard Arabic of statutes is not the Egyptian or Gulf dialect people actually type, and *legal* Arabic is a third register again — formal, archaic, full of terms of art. A model fluent in conversational Arabic can still miss a contract clause the way a fluent English speaker might fumble a 17th-century deed.

And — counterintuitively — the problem usually *isn't* missing content. We ran the same legal questions in both languages and Arabic surfaced **9× more primary law** than English; the catch was that those sources sit un-indexed on bare government servers, and retrieval kept mixing up *which country's* law it found. We wrote that up in full in [the Arabic legal AI gap](/blog/arabic-legal-ai) — the short version is that the hard part is retrieval and jurisdiction accuracy, not vocabulary. This post is about the layer above that: who's actually shipping products on top of this reality, and how far they've got.

Who's actually building it

More people than the English-language press thinks. Honestly mapped, by country:

**Saudi Arabia — the deepest market.** [Adel](https://tryadel.sa) is the most traction-proven Arabic legal AI we found: a consumer-and-pro app with 663 App Store ratings (4.6★), a claimed 500K downloads, 70,000+ Saudi legal documents, and real published pricing (SAR 149–199/month). [Shwra](https://shwra.ai) has the widest consumer distribution — iOS, Android *and* Huawei, ~1,400 App Store ratings — built as a hybrid: an AI assistant ('Mishir') that triages and then routes you to a licensed human lawyer. [Qaanoon](https://qaanoon.ai) is the free, no-friction Arabic chatbot for individuals. [Laika](https://laika.legal) and [Malakah](https://malakah.ai) round out a genuinely competitive Saudi field. Every one of them is Saudi-only.

**UAE.** [realLaw](https://reallaw.ai) is a consumer UAE app (free tier + AED 74/month) carrying a Dubai government AI certification. [Qanooni](https://qanooni.ai) ($2M pre-seed, Village Global) goes the B2B route with Word/Outlook integration. And the wildcard isn't a startup at all: the **UAE Ministry of Justice**, with vendor GenArabia, has deployed an Arabic legal-AI assistant covering 5,000+ pieces of legislation — as kiosks at court entrances. Free, government-run, and reshaping what citizens expect.

**Egypt.** [LegalMind](https://getlegalmind.com), 'made by Egyptian lawyers for Egyptian lawyers,' is the dedicated Egyptian-law tool — B2B, web-only, one country.

**Pan-Arab.** [Laiwyer.ai](https://laiwyer.ai) is the rare multi-country play — Qatar, UAE, KSA and Egypt, transparent pricing ($49–99/mo) — but it's a web-only research tool for lawyers, not a consumer app. And [Arabic.ai](https://arabic.ai/legal) (partnered with Jordan's Qistas) is the most technically serious of all: a genuinely *sovereign* Arabic-first LLM — its own models, 22 dialects plus MSA — not a wrapper on GPT. But it sells only to enterprises and governments, behind a procurement cycle. No consumer can touch it.

The four-corner test

Line them up against four questions a person — not a law firm — actually cares about:

Adel and Shwra ace 1, 2 and 4 — and fail 3 (Saudi only). Laiwyer aces 3 and 4 — and fails 1 and 2 (web, lawyers). Arabic.ai has the best 4 in the business — and fails 1, 2 and 3 for any individual. **Every serious player lands two or three corners. None lands all four.** That's not a knock on any of them; single-country depth is a perfectly good strategy. It's just where the open space is.

Every Arabic legal AI, compared

The full MENA/Arabic field we found, with the honest details — audience, platform, pricing, whether the Arabic is native, and whether it crosses borders. (We excluded Perle AI: despite the 'Arabic legal' framing, it's a data-annotation platform, not a legal tool.)

Read down the last two columns. The two 'yes-and-yes' rows that are also consumer + mobile? There's one — and it launched last week.

Three gaps nobody's closed

What we're doing about it

Disclosure, since this is our blog: we build HAQQ Legal AI, and we built it to aim at all four corners at once — consumer, mobile, multi-jurisdiction, native Arabic. Our engine, **Justinian**, handles Arabic right-to-left as a first-class language and is jurisdiction-aware across 80+ countries' legal systems, not one. The app is [live on the App Store](https://apps.apple.com/us/app/haqq-legal-ai/id6760749447) (Android is coming).

The honest caveat, same as our [global app teardown](/blog/ai-lawyer-app-landscape-2026): we're the newest name on this page. Adel has half a million downloads and we have a launch. Arabic.ai has a sovereign model with benchmark scores we respect and don't claim to beat. What we have is a bet on the one combination nobody else is making — and after mapping the field, we're more convinced the corner is real than that we're the only ones who'll ever stand in it.

Key takeaways

Sources & further reading