AI projects in the Kingdom are discussed in two languages, often in the same meeting. This glossary sets out 32 terms that come up when planning, buying or running AI, each with a widely used Arabic equivalent and a one-sentence explanation in plain words. Many teams also use the English word as it is. What matters is that everyone on a project means the same thing.
Models & training
| English | Arabic | In plain words |
|---|
| Large language model (LLM) | نموذج لغوي كبير | A model trained on very large amounts of text to read, write and answer in natural language. |
|---|
| Open-weight model | نموذج مفتوح الأوزان | A model whose trained weights are published, so it can run on infrastructure you control. |
|---|
| Token | الرمز (التوكن) | The small unit of text a model reads and writes, often part of a word; model limits and many prices are counted in tokens. |
|---|
| Context window | نافذة السياق | How much text a model can take into account at once, measured in tokens. |
|---|
| Prompt | الأمر النصي | The instructions and input given to a model for a task. |
|---|
| Fine-tuning | الضبط الدقيق | Further training an existing model on your own examples to adapt it to a specific task or style. |
|---|
| Hallucination | الهلوسة | A confident answer that is not supported by the sources or the facts. |
|---|
Retrieval & knowledge
| English | Arabic | In plain words |
|---|
| Retrieval-augmented generation (RAG) | التوليد المعزز بالاسترجاع | The system first finds the relevant passages in your documents, then the model answers from them. |
|---|
| Embeddings | التضمينات | Numbers that represent the meaning of text, so search can match meaning rather than exact words. |
|---|
| Vector database | قاعدة بيانات متجهية | A database built to store embeddings and find the closest matches quickly. |
|---|
| Semantic search | البحث الدلالي | Search by meaning, so a question about “annual leave” can find a policy titled “holiday entitlement”. |
|---|
| Chunking | تجزئة النصوص | Splitting long documents into smaller passages so the right section can be found and quoted. |
|---|
| Grounding | الإسناد إلى المصادر | Tying each answer to source passages the reader can open and check. |
|---|
| Optical character recognition (OCR) | التعرّف الضوئي على الحروف | Turning scanned pages and photos into searchable text; Arabic needs OCR that handles its connected script. |
|---|
Agents & automation
| English | Arabic | In plain words |
|---|
| AI agent | الوكيل الذكي | A system that uses a model to plan steps and use tools, such as looking up a record, to complete a task. |
|---|
| Tool calling | استدعاء الأدوات | A model asking another system to do something specific, such as run a search or update a record. |
|---|
| Human-in-the-loop | الإنسان ضمن الحلقة | A design in which a person reviews or approves certain steps before the system continues. |
|---|
| Escalation | التصعيد | Handing a case to a person when it falls outside what the system is allowed or able to handle. |
|---|
| Intelligent document processing | المعالجة الذكية للمستندات | Reading documents such as invoices, extracting their fields and checking them against your records. |
|---|
Security & governance
| English | Arabic | In plain words |
|---|
| Prompt injection | حقن الأوامر | Instructions hidden in content, such as an e-mail or a document, that try to make a model ignore its rules. |
|---|
| Guardrails | الضوابط الوقائية | Rules and checks that limit what a model or agent may say or do. |
|---|
| Least privilege | مبدأ أقل صلاحية | Giving a user or agent only the access a task needs, and nothing more. |
|---|
| Audit trail | سجل التدقيق | A record of who asked what, which sources were used and which actions were taken. |
|---|
| Data residency | إقامة البيانات | Where data is physically stored and processed, for example inside the Kingdom. |
|---|
| Data classification | تصنيف البيانات | Labelling data by sensitivity, which decides where it may be stored and which systems may use it. |
|---|
| Personal data | البيانات الشخصية | Any information that can identify a person; in the Kingdom its handling is governed by the Personal Data Protection Law (PDPL). |
|---|
Operations
| English | Arabic | In plain words |
|---|
| Inference | الاستدلال | Running a trained model to produce an answer; the everyday workload once a system is live. |
|---|
| GPU | المعالج الرسومي | The processor most AI models run on; its memory largely decides which models fit. |
|---|
| Quantisation | التكميم | Storing a model’s numbers at lower precision so it needs less memory, usually with a small loss of quality. |
|---|
| Latency | زمن الاستجابة | The time between a request and its answer. |
|---|
| Evaluation set | مجموعة التقييم | A fixed set of real questions with agreed good answers, used to measure quality before and after any change. |
|---|
| Model drift | انحراف النموذج | A gradual change in quality as real questions and data move away from what the system was tested on. |
|---|