Resources and learning materials

Materials for reviewers, partners, and replication teams

Project files remain marked pending until verified documents are uploaded. The external library below points to trusted AI ethics, AI literacy, digital inclusion, and Arabic-language AI resources.

Pending upload

Open link

Project brief

A concise overview of the initiative, users, partners, and replication pathway.

One-page concept note

Short application-ready framing for reviewers and partners.

AI literacy guide

Training material for safe and practical AI use by local civic actors.

Safe AI use checklist

Privacy, verification, misinformation, and escalation checklist.

Methodology note

Co-design, training, safeguards, and tool-scoping methodology.

Replication brief

Southern Mediterranean adaptation through EuroMedAI.

UNESCO Recommendation on the Ethics of AI

Global policy reference for human rights, gender equality, privacy, transparency, and responsible AI governance.

UNESCO AI and Education: Guidance for Policy-Makers

UNESCO publication page for the policy guide explaining AI, machine learning, policy choices, and education risks in accessible language.

ESCWA Artificial Intelligence Futures for the Arab Region

Regional report on AI opportunities, governance risks, data, ethics, inequalities, and Arabic-language AI development.

UN ESCWA

ESCWA Learn

Arabic e-learning portal with regional courses on digital transformation, development, public leadership, and emerging technologies.

ITU Academy

Capacity-building platform for ICT policy, digital inclusion, cybersecurity, AI governance, and emerging technologies.

Google Skillshop

Google learning portal with Arabic interface options for practical digital skills and product-focused learning paths.

Google Machine Learning Crash Course

Arabic-accessible Google Developers material on machine learning foundations, datasets, evaluation, and model behavior.

Microsoft Azure AI Fundamentals

Arabic Microsoft Learn certification pathway for AI concepts, machine learning basics, generative AI workloads, and responsible AI principles.

Microsoft Learn

MBZUAI Institute of Foundation Models

Research hub for Arabic and multilingual foundation models, including the JAIS series and responsible generative AI research.

KAUST Generative AI Research

KAUST research page on generative AI, Arabic-language AI directions, and applied AI work from the region.

A glossary of AI terms in plain language - and in Darija

Speaking AI, locally

Replicating this model means making AI vocabulary legible to civic actors. These are the terms we use, with Arabic and Darija equivalents where they exist.

All terms

Search a term...

No terms match your search.

In Darija

جواب مخترع

When an AI produces a confident answer that is not grounded in real sources - it may invent a law, a citation, or a fact. AMSA mitigates this by grounding every answer in a retrieved corpus.

عندما ينتج الذكاء الاصطناعي جواباً واثقاً غير مستند إلى مصادر حقيقية - قد يخترع قانوناً أو مرجعاً. تحدّ أمسا من ذلك بربط كل جواب بمصدر مُسترجَع.

Retrieval-Augmented Generation (RAG)

التوليد المعزز بالاسترجاع

An architecture where the AI first retrieves relevant documents from a trusted corpus, then writes an answer based only on what it found - rather than relying on memory alone. This is what keeps AMSA grounded.

بنية يقوم فيها النظام أولاً باسترجاع وثائق ذات صلة من مدوّنة موثوقة، ثم يصوغ الجواب اعتماداً على ما وجده فقط، بدل الاعتماد على الذاكرة وحدها.

The curated collection of trusted documents an AI is allowed to draw from. AMSA's corpus is official Moroccan civic sources: the Constitution, Organic Laws 111/112/113.14, and DGCT and IEECAG materials.

المجموعة المنتقاة من الوثائق الموثوقة التي يُسمح للنظام بالاعتماد عليها. مدوّنة أمسا هي مصادر مدنية مغربية رسمية.

السؤال اللي كتكتب

The text you type to the AI - your question or instruction. Clear prompts lead to more useful answers.

النص الذي تكتبه للنظام - سؤالك أو تعليمك. الطلبات الواضحة تؤدي إلى أجوبة أنفع.

السند القانوني

The reference to the exact article or document an answer is based on. Every AMSA answer links its claims to a source so you can verify.

الإشارة إلى المادة أو الوثيقة التي يستند إليها الجواب. كل جواب من أمسا يربط ما يقوله بمصدر يمكنك التحقق منه.

تحويل لخبير

The rule that sends a question to a human expert when it needs binding legal counsel or touches a dispute. AMSA prepares; it does not substitute for a lawyer or authority.

القاعدة التي تحوّل السؤال إلى خبير بشري عندما يتطلب استشارة قانونية ملزمة. أمسا تُحضّر ولا تعوّض المحامي أو السلطة.

Large Language Model (LLM)

النموذج اللغوي الكبير

The kind of AI that understands and generates text. AMSA uses one, but constrains it with retrieval and citations so it stays grounded in real sources.

نوع الذكاء الاصطناعي الذي يفهم النص ويولّده. تستعمله أمسا لكن تقيّده بالاسترجاع والإحالات ليبقى مستنداً إلى مصادر حقيقية.

الإنسان في الحلقة

A safeguard where people review and oversee the AI rather than letting it act alone. Flagged answers and sensitive cases are routed to the project team.

ضمانة يشرف فيها البشر على النظام بدل تركه يعمل وحده. الأجوبة المُبلَّغ عنها والحالات الحساسة تُحوَّل لفريق المشروع.

إخفاء الهوية

Storing data without linking it to a real identity. AMSA ties conversations only to an opaque browser-generated ID, never to your name or phone number.

تخزين البيانات دون ربطها بهوية حقيقية. تربط أمسا المحادثات بمعرّف مبهم فقط، لا باسمك أو رقم هاتفك.