Qatar data stays in Qatar. Your AI keeps working.

Protecto masks Arabic and English personal data before any model sees it, and runs on-premises, in a private cloud, or air-gapped.

Qatar sovereignty rulesIllustrative

Protectoapplies the rule that fits each unit

Law No. 13 of 2016

Health and family notes need permission (Art. 16)

Held back

QFC Regulations 2021

Data leaving the QFC needs a basis (Arts. 23, 24)

Tokens only

QCB Cloud Regulation

Personal and financial data processed in Qatar only (cl. 21.4)

Model in Qatar

Protecto sits between your data and every model.

Trusted by regulated enterprises & agentic platforms

Case study result

Bank Muscat, Arabic and English data

Read the case study →

99%

Recall

96%

Precision

4 weeks

To a working proof

The problem

One assistant for every team means three sets of rules.

A Doha financial group builds one AI assistant for all its teams. The bank’s customer files must be processed in Qatar. The advisory arm in the Qatar Financial Centre follows its own transfer rules. The home-services arm follows the national law, where a note about a surgery or a divorce needs a government permission before anyone processes it. The assistant reads all of them.

The need

Your team wants one assistant

Support tickets, claims notes, and customer files are where AI saves the most time. They hold names, ID numbers, and personal details, in Arabic and English.

The challenge

The rule depends on the unit

A bank using cloud must process personal and financial information within Qatar only (QCB Cloud Computing Regulation, cl. 21.4). A QFC firm needs a basis to send data out (QFC Regulations, Arts. 23 and 24). Everyone else follows Law No. 13 of 2016, which has no hosting rule.

The gap

A Doha data center settles one rule

Hosting the model in Qatar helps the bank. It does not give permission for health or family data, and a log or a support engineer abroad can still reach a copy. DLP that blanks every value breaks the AI.

The risk

A divorce note is regulated data

Law No. 13 treats health, children, marital relations, and similar data as special-nature data that needs permission (Art. 16). Fines reach QR 5 million (Art. 24). Old tickets hold these notes in free text.

How it works

Real values stay in Qatar. Only tokens cross.

A Doha bank’s support agent reads a ticket in Arabic and English. Protecto runs in your Qatar environment and swaps the sensitive values for tokens. A model you host gets tokens, and so does any outside model your reviewers approve. Real values never leave. Each unit gets its own policy, so a bank, a QFC firm, and any other company can each follow their own rule.

Real timeProtects model input, retrieval, logs, tool calls, and agent memory as data moves, not just stored files.

Your Qatar environment

Your apps and data

Chat appsAI agentsMCP serversDocuments

Real values

خالد المري29001234567

Protecto

Detect, tokenize, apply policy

Token vaultAudit log

Tokens only

[PERSON_9e2][ID_5b2]

Model you host

Receives tokens

Ollama

DeepSeek

Moonshot AI

The answer comes back as tokens. Protecto restores real values only for the roles your policy names.

Only tokens cross

[PERSON_9e2][ID_5b2]

Qatar border

Optional, outside Qatar

Any AI model

OpenAI

Anthropic

Google

xAI

Sees tokens, never real values.

Tokens are still personal data, and whoever holds the mapping can reverse them. An outside path still needs its own basis under your rulebook. An air-gapped setup has no outside path.

Choose your isolation

On-premises

Runs on your servers, in your Qatar data center.

Private cloud

Runs in your cloud account, in a region you choose.

Air-gapped

No outside link. The model runs inside the same boundary.

Connect it to your agents

API or proxy

Call Protecto before the model, or route model traffic through it.

MCP and tool calls

Protecto sits in the path of tool calls, so agents keep their workflows.

Start with one flow

One use case, one Arabic data sample. Add flows as reviewers sign off.

Before the AI reads it

Years of tickets hold notes the law treats as special.

Free text is where special-nature data hides. A clerk typed that a customer was in the hospital, or that a plan moved to an ex-wife after a divorce. Protecto flags these notes before they reach an index or a model, and your policy decides what happens next.

Ticket archive scanIllustrative

Note in the ticketWhat it isWhat Protecto does

Please move the plan to the new villa in Lusail.

Personal dataAddress

TokenizedSame token every time

Customer was in the hospital for two weeks, so the visit was moved.

Special-nature dataHealth (Law 13, Art. 16)

Held backUnless your policy releases it

Account is now in his ex-wife's name after the divorce.

Special-nature dataMarital relations (Art. 16)

Held backUnless your policy releases it

Parent says the 9-year-old's inhaler order is late.

Special-nature dataChildren and health (Art. 16)

Held backUnless your policy releases it

Every release is logged, so your team can show who saw which values, and when.

For your reviewers

Seven rules behind the three rulebooks, and what Protecto does about each.

Fines reach QR 5 million under Law No. 13 of 2016 (Art. 24) and USD 1.5 million per provision under the QFC Regulations (Art. 36(3)). References are to Law No. 13 of 2016, the QFC Data Protection Regulations 2021, and the QCB Cloud Computing Regulation (in force 15 April 2024).

Rule	Requirement	Protecto solution
Which rulebook appliesLaw 13, Art. 2 QFC Regs, Arts. 2, 7 QCB Cloud Reg, cl. 5.1	Law No. 13 covers electronic processing of personal data. The QFC Regulations cover firms registered in the QFC, and others that process through them on an ongoing basis. The QCB regulation covers cloud use by QCB-licensed entities. One group can face all three.	Protecto applies a policy per use case, so each unit gets the rule that fits it. The checkpoint is the same.
Where data is processedQCB Cloud Reg cl. 21.4, 21.5 Law 13, Art. 15	A QCB entity must ensure personal and financial information is processed within Qatar only, and get QCB approval before a cloud arrangement. Law 13 has no hosting rule.	Protecto runs on-premises, in your private cloud, or air-gapped, with the model inside the same boundary. Real values, mappings, and keys stay in your environment.
Transfers out of the QFCQFC Regs Arts. 23, 24	A recipient outside the QFC needs an adequate jurisdiction, safeguards such as standard clauses, a listed derogation, or a permit.	Protecto sends tokens, not real values. Tokens are still personal data, so the basis for the path stays your decision.
Special-nature and sensitive dataLaw 13, Arts. 16, 24 QFC Regs, Art. 12	Law 13: health, children, marital relations, ethnic origin, religious creeds, and criminal offenses need the Competent Department's permission. QFC: sensitive data is barred unless a listed ground applies.	Protecto flags these notes in Arabic and English text and holds them back from the model, the index, and the logs unless your policy releases them. Test it on your own tickets.
Security and pseudonymizationLaw 13, Arts. 8(3), 13 QFC Regs, Arts. 26, 29	Precautions that fit the nature and importance of the data. The QFC Regulations name de-identification or encryption, and access only for roles that need the data.	Tokens replace real values before data reaches a model, and map back only under policy. The QFC Data Protection Office says de-identified data is still personal data.
Processors, records, and breachesLaw 13, Arts. 11(8), 12, 14 QFC Regs, Arts. 28, 30, 31	Verify processors, keep records, and report breaches. QFC: to the Data Protection Office within 72 hours where feasible. Law 13: tell the person and the Competent Department if serious damage is possible.	Real values stay out of models, vector stores, and logs, so less is exposed. The audit log shows what each recipient received.
Review before launchLaw 13, Art. 11(1) QFC Regs, Art. 27	Review privacy measures before new processing. The QFC requires an impact assessment for high-risk processing, including large-scale sensitive data.	The scan shows which sensitive data a flow touches, so your review starts from facts.

Beyond these three: health, telecom, and government rules can add duties for some teams. Check your sector regulator.

The whitepaper follows one agent through every rulebook and every data path, with what each recipient may receive.

Case study

How Bank Muscat got to a working proof in four weeks.

Financial services

Data residency rules blocked cloud AI. A mix of Arabic and English data made it harder.

Bank Muscat wanted to adopt AI but could not send customer data to top cloud models, and its Arabic, English, and mixed-language text made sensitive data hard to find. With Protecto, sensitive data is masked by default and unmasked only under policy. The team had a working proof of concept in four weeks.

99%

Recall on sensitive data, including Arabic edge cases

96%

Precision, so useful text is not blanked out

4 weeks

From start to a working proof of concept

Bring one use case and one Arabic data sample. We will show it protected, inside your network.

Resources

Go deeper.

Whitepaper · 30 pages

Qatar Data Protection for Enterprise AI and Agents

Map the applicable regime, control disclosure, and keep agent actions useful. One Doha support agent, followed through Law No. 13 of 2016 and the QFC Regulations.

Blog

What Is a Data Privacy Vault?

How vault-based tokenization keeps sensitive data out of AI workflows.

Read the post

Guide

How to Ensure Data Privacy with AI

A step-by-step guide to protecting personal data in AI projects.

Read the guide

Industry

Sovereign AI for banks

How banks use leading LLMs while keeping sensitive data inside their boundary.

See the solution

FAQ

Questions security and compliance teams ask first.

Does Qatar law say personal data must stay in Qatar?

Not in Law No. 13 of 2016. The law has no hosting rule, and Article 15 bars a controller from limiting cross-border data flow unless the processing breaks the law or may cause serious damage. Other rules differ. A bank or other entity licensed by the Qatar Central Bank (QCB) must make sure personal and financial information is processed within Qatar only when it uses cloud (Cloud Computing Regulation, cl. 21.4), and a firm in the Qatar Financial Centre needs a basis to send data outside the QFC (Arts. 23 and 24). Sector rules can add more. Confirm your position with counsel and your regulator.

Which Qatar rulebook applies to us?

It depends on the entity and the activity. The QFC Regulations apply to firms incorporated or registered in the QFC, and to others that process personal data through a QFC firm on an ongoing basis, for that processing only (Art. 7). The QCB regulation covers cloud use by QCB-licensed entities. Law No. 13 of 2016 covers electronic processing of personal data more widely. One group can fall under all three. Name the entity and the rulebook for each AI flow before you design it.

If our model runs in Doha, are we done?

No. Where the model runs is one fact. Model inputs, logs, vector stores, and vendor support can still carry copies to other places, and hosting does not give you the permission the national law asks for on health and family data. List every path, who can reach it, and where they work. Protecto swaps sensitive values for tokens before the data reaches any of them.

Can a bank use a model outside Qatar?

Clause 21.4 of the QCB Cloud Computing Regulation says an entity must ensure personal and financial information is processed within Qatar only, and clause 21.5 requires QCB approval before a cloud arrangement. Tokens are pseudonymized, not anonymized, so whether tokens count as personal information under 21.4 is a question for your QCB contact. If the answer is not clear, run the model inside Qatar, for example air-gapped.

What is special-nature data under Law No. 13?

Article 16 lists ethnic origin, children, health, physical or psychological condition, religious creeds, marital relations, and criminal offenses. Processing needs permission from the Competent Department, and a violation can carry a fine up to QR 5 million (Art. 24). Free-text notes such as a hospital stay or a divorce count. Protecto flags these notes so your policy can hold them back from the model, the index, and the logs.

Can Protecto run with no internet connection?

Yes. Air-gapped is one of the ways to deploy Protecto, alongside on-premises and a private cloud. In an air-gapped setup the AI model has to run inside the same boundary, so you host it yourself, for example with Ollama. Nothing in the flow needs an outside connection.

How hard is it to add Protecto to our existing apps?

Protecto is an API product. Your team calls the API before the model, or routes model traffic through Protecto as a proxy, and your existing app logic stays as it is. There is no new interface to roll out. Start with one AI flow and one data sample, then add more flows as reviewers sign off.

Does Protecto work on Arabic text and Qatar ID numbers?

Protecto detects 50+ entity types out of the box, covering personal data (PII), health data (PHI), and payment card data (PCI). It also supports custom entities, so you can add identifiers specific to your business, such as a Qatar ID format.

Will masking make the AI's answers worse?

Basic redaction can, because a blank placeholder tells the model nothing. Protecto uses consistent, format-preserving tokens, so the same person maps to the same token and the model can still tell who is who. Test it on your own data before relying on any figure.

Is tokenized data still personal data?

Tokenization is pseudonymization, not anonymization. Whoever holds the token mapping can reverse a token, so treat tokenized data as personal data unless counsel advises otherwise. Law No. 13 counts data as personal when a person can reasonably be identified, including by combining it with other data (Art. 1). The QFC Regulations name de-identification as a security measure (Art. 29), and the QFC Data Protection Office says de-identified data is still personal data.

Try it on your own Qatar data.

Bring one AI use case and one Arabic data sample. We will show where its data goes and what each model, log, and support team receives.