Face Recognition + PDPL: What You Can and Cannot Do in Saudi Arabia

What PDPL says about biometric data

The Personal Data Protection Law treats biometric data — including face templates and facial vectors — as a special category requiring stronger protection than ordinary personal data. The implementing regulations (issued by SDAIA) set out:

1. Lawful basis — explicit consent is the default; legitimate interest is permitted only with documented justification.
2. DPIA requirement — face recognition triggers a mandatory Data Protection Impact Assessment.
3. Cross-border rules — biometric data is among the categories most tightly bound by Article 29 cross-border rules.
4. Retention — must be the minimum necessary; default postures of “keep forever” will fail.

Anchor in the PDPL glossary and the PDPL compliance video analytics checklist.

What you can do — three permitted use cases

1. Workplace access control with explicit, documented employee consent and an alternative method (badge or PIN) for those who refuse.
2. High-security site access with a documented legitimate-interest test, often paired with consent. Common at Aramco and military-adjacent sites.
3. Safety alerting — recognising specific authorised personnel in restricted zones, again with the legitimate-interest test and DPIA.

For the underlying solution see the face recognition solution and the access control solution.

What you cannot do — three common mistakes

1. Continuous covert face recognition of customers in public retail spaces without consent infrastructure. Not defensible under 2026 PDPL.
2. Cross-border face template export to a vendor’s overseas data centre without an Article 29 mechanism in place.
3. “Just in case” indefinite retention of face vectors. Retention must be tied to a documented purpose.

A vendor proposal that requires any of these three is a red flag.

The DPIA — what auditors expect

A defensible 2026 DPIA for face recognition contains:

1. Purpose specification — what business outcome the system enables.
2. Necessity test — why face recognition is required versus alternatives (badge, PIN, fingerprint).
3. Proportionality test — scope, retention, access controls.
4. Risk register — bias risk, false-match risk, function-creep risk.
5. Mitigation plan — minimum necessary data, KSA-resident processing, retention schedule.
6. DPO sign-off with named signatory.

This anchors in the trust / compliance overview and the security overview.

Cross-border rules under Article 29

Article 29 binds the export of personal data — and especially biometric data — outside the Kingdom. Three permitted patterns in 2026:

1. No transfer. Face templates stored and processed entirely inside CST-licensed data residency regions. The defensible default.
2. Adequacy decision — the destination country has a SDAIA adequacy decision. Few qualify in 2026.
3. Specific safeguards — explicit consent, contractual safeguards, or specific necessity under Article 29 exceptions.

For detail see the PDPL compliance video analytics answer.

Employee monitoring — the labour-law overlay

PDPL is not the only constraint. Employee monitoring also touches:

• Saudi Labour Law — monitoring must be proportionate and disclosed.
• NCA Essential Cybersecurity Controls — see the NCA-ECC glossary.
• Sectoral rules — Aramco, SABIC and other operating organisations layer their own contractor HSE manuals on top.

The defensible posture combines explicit employee consent, a documented purpose, and a clear opt-out path that does not penalise the worker.

Bias and false-match risk

Biometric systems exhibit demographic performance variance. KSA construction workforces are unusually diverse — South Asian, North African, Filipino, Saudi national. A 2026-grade vendor:

1. Discloses per-demographic precision/recall on a held-out validation set.
2. Tracks drift weekly, with retraining triggered below an operational threshold.
3. Provides a manual override path so a falsely rejected worker can still enter through a secondary verification.

Without these, the system’s first false-match incident will collapse trust.

Retention — what 2026 looks like

A defensible retention schedule:

Data class	Retention
Live face template (active access)	While employment / access continues
Match log (event metadata)	30 days default, 90 days if linked to incident
Face vector backup	None — re-enrol if needed
Face image archive	Not retained beyond enrolment

Anchor in the data residency posture.

Vendor selection filter

Three filters that trim the candidate list:

1. Saudi-resident processing confirmed in writing.
2. DPIA template provided as part of the proposal, not as a billable add-on.
3. Bias and demographic performance disclosure with the proposal.

For the broader procurement view see the top 10 platforms shortlist and the IKTVA Saudi-Made AI vendors guide.

What to expect in an SDAIA inspection

A 2026 SDAIA inspection of a face-recognition deployment typically looks for:

1. Lawful basis register, signed by the DPO.
2. DPIA on file, with named signatory.
3. Consent records (where applicable) with timestamps.
4. Retention schedule applied — random sample check.
5. Cross-border posture documented.
6. Incident response plan for biometric breaches.

If any of these is missing, expect a remediation order with a 30–90 day deadline [VERIFY-SME].

Common deployment mistakes

1. Skipping the DPIA — the most common failure mode.
2. No alternative method — workers cannot decline face enrolment.
3. Cross-border processing without an Article 29 mechanism.
4. Indefinite retention — “forever” is not a retention period.

Next steps

If you are scoping or auditing a face recognition deployment in Saudi Arabia, start with the face recognition solution, the access control solution and the PDPL compliance checklist. Cross-reference the PDPL glossary entry and the trust / compliance overview.

Book a privacy-side scoping call and we will produce a DPIA template and lawful-basis register tailored to your deployment within 10 working days.

React to this article

🔥...💡...🤔...👏...❤️...