Research, method note, November 2026
Why Arabic names break sanctions screening
One Arabic name can have hundreds of valid Latin spellings. Where the variation comes from, why fuzzy matching does not solve it, and five habits that do most of the work.
Arabic names are written in Arabic script. Every Latin spelling is somebody's transliteration, and there is no single system. A passport officer, a bank clerk, a journalist and a sanctions authority may each spell the same person differently, and all of them are right.
One name, many spellings
عبد الرحمن: Abdelrahman, Abdel Rahman, Abdul Rahman, Abd al-Rahman, Abdulrahman, Abdurrahman
عثمان: Osman, Othman, Uthman
الطيب: El Tayeb, Eltayeb, Al-Tayeb, Altayib, El-Tayib
A typical Sudanese name has three or four parts: a given name, then the father's and grandfather's names. If each part has four to six common spellings, the full name has several hundred valid Latin forms.
Where the variation comes from
| Source of variation | Examples |
|---|---|
| Vowels. Short vowels are not written in Arabic, so they are guessed | Mohamed / Muhammad; Hussein / Hussain / Husayn |
| The definite article | al-; el-; Al; El; attached or separate; ar-Rahman / al-Rahman |
| The letter ق, pronounced like a hard "g" in Sudan | Gedaref / Qadarif; Gurashi / Qurashi |
| The letter ج, "j" in Sudan and "g" in Egypt | Jamal / Gamal; Jaafar / Gaafar |
| The letter ث | Osman / Othman / Uthman |
| Compound names with عبد | Abd; Abdel; Abdul; Abdal; Abdu, joined or spaced |
| Doubled letters and endings | Mohamed / Mohammed; Fatima / Fatimah; Hamza / Hamzah |
| How many parts are given | Two names on a business card, three on an invoice, four on a passport |
| Nicknames | Prominent figures are often known publicly by a nickname that appears on no official document |
Why fuzzy matching does not solve it
Most screening tools tolerate a few changed letters. That catches typing errors. It does not reliably catch a different transliteration system, where "Osman" and "Uthman" share three letters out of six, or a name that is split differently, where "Abdelrahman" is one token and "Abd al Rahman" is three. Tighten the tolerance and you miss true matches. Loosen it and a common name such as Mohamed Ahmed returns hundreds of false ones, which a tired reviewer then clears in bulk.
Five habits that fix most of it
- Always collect the name in Arabic script. Ask the counterparty for it, or take it from their registration papers, stamp or letterhead. Arabic is the fixed point; every Latin form is a derivative.
- Screen on the Arabic as well. The UN, US, UK and EU lists carry names in original script or in alias fields for many Arabic-speaking designees. Those fields are rarely searched.
- Generate the variants deliberately. Produce the common spellings of each part and screen the combinations, instead of relying on a tool's tolerance setting.
- Collect a second and third identifier. Father's and grandfather's names, date or place of birth, company role, registration number. A name match alone is not an identification, and a name mismatch alone is not a clearance.
- Record what you searched. List the spellings in the file. When a new designation appears next year, you will know whether your earlier search would have caught it.
Common names cut both ways. The same problem that hides true matches also produces false ones. Before treating a hit as a finding about your counterparty, establish that it is the same person. Attributing a sanctioned person's record to someone who merely shares a name does real harm.
Larva Intelligence screens on Arabic script and generated variants as standard. The method