Legal notice
This page identifies who operates ArabicKeyboard.ai and how it is published, as required for pre-launch transparency under Swiss and EU law.
Operator
Mustapha El Amrani
Oerlikonerstrasse 42
8057 Zurich, Switzerland
General contact: contact@arabickeyboard.ai
Privacy requests: privacy@arabickeyboard.ai
Hosting
This site is hosted on Vercel (Vercel Inc., vercel.com). Vercel serves the site's pages and static assets and processes ordinary web request data in the course of delivering them — see the privacy page for what that involves.
Copyright
© 2026 Mustapha El Amrani. The text, design, and software of this site may not be reproduced without permission, except as permitted by applicable law.
Liability
This site and its typing tools are provided as-is, without warranty of uninterrupted availability or fitness for a particular purpose. Nothing in this notice excludes or limits any liability that cannot lawfully be excluded or limited under applicable Swiss or EU law, including liability for death, personal injury, or fraud.
External destinations
Google Search and YouTube open only when you deliberately click the corresponding key on the virtual keyboard, and only after you confirm a dialog that names exactly what is about to happen. Once that new tab opens, Google's or YouTube's own terms and privacy policy govern how they handle what you sent them — this site has no further involvement or control.
Open-source components
Local Arabic spellchecking on this site (see the online Arabic keyboard) is built on Hunspell (v1.7.3, tri-licensed GPL-2.0-or-later / LGPL-2.1-or-later / MPL-1.1) and the Ayaspell Arabic dictionary, distributed via LibreOffice/dictionaries (same tri-license). Both are compiled/shipped unmodified from their upstream source. The exact versions, checksums, compile command, applicable license terms, and the complete, self-hosted copy of the Hunspell source this site actually compiles are documented in this project's own spellchecking source-and-license notice.
The local PNG text-card export on the Arabic keyboard pages draws with three Arabic fonts served from this site, each redistributed unmodified under the SIL Open Font License 1.1: Scheherazade New v4.500 (© 1994–2026 SIL Global, Reserved Font Names “Scheherazade” and “SIL”); Rakkas (© 2016 Zeynep Akay); and Noto Kufi Arabic v2.110 (© 2022 The Noto Project Authors). The complete licence text for each font, including its own copyright notice, is served alongside the font file itself — Scheherazade New, Rakkas and Noto Kufi Arabic. Exact upstream versions and file checksums are recorded in the text-card font source-and-licence notice this site serves alongside the fonts themselves. The font authors do not endorse this product, and the font files are not sold or offered separately from it.
The Arabic Check games use two further fonts served from this site, each redistributed unmodified under the SIL Open Font License 1.1: Readex Pro (© 2018 The Readex Pro Project Authors, Reserved Font Name “RevReading Lexend”) and Aref Ruqaa (© 2015–2020 The Aref Ruqaa Project Authors, Reserved Font Name EURM10), both taken from the google/fonts repository. Their complete licence texts are served alongside the fonts — Readex Pro and Aref Ruqaa. The font authors do not endorse this product, and the font files are not sold or offered separately from it.
The 3D hands in the touch-typing course are drawn with three.js (v0.170.0, MIT License, © 2010–2024 three.js authors) using the unmodified “generic-hand” models from the WebXR Input Profiles assets (MIT License, © 2019 Amazon). Both are downloaded from this site only when the course comes into view with the hands switched on (the default); the complete MIT licence texts are shown in the course's own “credits and licences” section.
The Latin-to-Arabic phonetic conversion feature draws on three data sources, each Wiktionary-derived and licensed CC BY-SA 4.0 / GFDL 1.1 or later (commercial reuse permitted, with attribution and a share-alike obligation on the derived data): WikiPron pronunciation data; Kaikki/Wiktextract structured data (Tatu Ylonen, “Wiktextract: Wiktionary as Machine-Readable Structured Data,” LREC 2022); and word-frequency data from the wordfreq project (Robyn Speer, licensed CC BY-SA 4.0), built from Wikipedia, OPUS OpenSubtitles, NewsCrawl, GlobalVoices, the OSCAR web crawl, and pre-2023 Twitter. None of these sources' own text is shown on this site — each is used only to help rank candidate Arabic spellings a visitor may choose from. Full source, revision, and license details are recorded in this project's own phonetic-conversion source-and-license notice.
Updated:
