Skip to content
All insights

Resource7 min readAahlan team

An Arabic–English glossary of AI terms

Thirty-two terms you will hear when planning, buying or running AI, each with its Arabic equivalent and a one-line explanation in plain words.

  • Glossary
  • Arabic AI
  • Fundamentals

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

EnglishArabicIn plain words
Large language model (LLM)نموذج لغوي كبيرA model trained on very large amounts of text to read, write and answer in natural language.
Open-weight modelنموذج مفتوح الأوزانA model whose trained weights are published, so it can run on infrastructure you control.
Tokenالرمز (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 windowنافذة السياقHow much text a model can take into account at once, measured in tokens.
Promptالأمر النصيThe instructions and input given to a model for a task.
Fine-tuningالضبط الدقيقFurther training an existing model on your own examples to adapt it to a specific task or style.
HallucinationالهلوسةA confident answer that is not supported by the sources or the facts.

Retrieval & knowledge

EnglishArabicIn plain words
Retrieval-augmented generation (RAG)التوليد المعزز بالاسترجاعThe system first finds the relevant passages in your documents, then the model answers from them.
EmbeddingsالتضميناتNumbers that represent the meaning of text, so search can match meaning rather than exact words.
Vector databaseقاعدة بيانات متجهيةA database built to store embeddings and find the closest matches quickly.
Semantic searchالبحث الدلاليSearch by meaning, so a question about “annual leave” can find a policy titled “holiday entitlement”.
Chunkingتجزئة النصوصSplitting long documents into smaller passages so the right section can be found and quoted.
Groundingالإسناد إلى المصادرTying each answer to source passages the reader can open and check.
Optical character recognition (OCR)التعرّف الضوئي على الحروفTurning scanned pages and photos into searchable text; Arabic needs OCR that handles its connected script.

Agents & automation

EnglishArabicIn plain words
AI agentالوكيل الذكيA system that uses a model to plan steps and use tools, such as looking up a record, to complete a task.
Tool callingاستدعاء الأدواتA model asking another system to do something specific, such as run a search or update a record.
Human-in-the-loopالإنسان ضمن الحلقةA design in which a person reviews or approves certain steps before the system continues.
EscalationالتصعيدHanding a case to a person when it falls outside what the system is allowed or able to handle.
Intelligent document processingالمعالجة الذكية للمستنداتReading documents such as invoices, extracting their fields and checking them against your records.

Security & governance

EnglishArabicIn plain words
Prompt injectionحقن الأوامرInstructions hidden in content, such as an e-mail or a document, that try to make a model ignore its rules.
Guardrailsالضوابط الوقائيةRules and checks that limit what a model or agent may say or do.
Least privilegeمبدأ أقل صلاحيةGiving a user or agent only the access a task needs, and nothing more.
Audit trailسجل التدقيقA record of who asked what, which sources were used and which actions were taken.
Data residencyإقامة البياناتWhere data is physically stored and processed, for example inside the Kingdom.
Data classificationتصنيف البياناتLabelling data by sensitivity, which decides where it may be stored and which systems may use it.
Personal dataالبيانات الشخصيةAny information that can identify a person; in the Kingdom its handling is governed by the Personal Data Protection Law (PDPL).

Operations

EnglishArabicIn plain words
InferenceالاستدلالRunning a trained model to produce an answer; the everyday workload once a system is live.
GPUالمعالج الرسوميThe processor most AI models run on; its memory largely decides which models fit.
QuantisationالتكميمStoring a model’s numbers at lower precision so it needs less memory, usually with a small loss of quality.
Latencyزمن الاستجابةThe time between a request and its answer.
Evaluation setمجموعة التقييمA fixed set of real questions with agreed good answers, used to measure quality before and after any change.
Model driftانحراف النموذجA gradual change in quality as real questions and data move away from what the system was tested on.

Want to talk this through?

We are happy to discuss how this applies to your organisation — no sales script.

Newsletter

Practical AI, in your inbox.

Real use cases, Saudi regulations and what works in Arabic — once or twice a month. No spam; unsubscribe any time.

We’ll e-mail you to confirm. Unsubscribe any time — see our privacy policy.