AI & personal dataUpdated 22 November 20269 min read

Facial Recognition and Biometrics Under the DPDP Act

Face recognition, fingerprint attendance and other biometrics are personal data under the DPDP Act. How to choose a basis, protect templates and avoid over-collection.

By Suhas Bhairav · Reviewed against the DPDP Rules, 2025 (G.S.R. 846(E), 13 Nov 2025)

Special biometric category?
No — ordinary personal data
Customer use basis
Usually specific consent (s.6)
Employee use basis
Possibly s.7(i), if proportionate
Security baseline
Rule 6(1)(a)–(g)
Max penalty (security failure)
Up to ₹250 crore
Applies from
13 May 2027

Quick answer

The DPDP Act has no special category for biometrics, but facial images, face templates and fingerprints are personal data, and their misuse is hard to undo. From 13 May 2027 any facial recognition or biometric system needs a lawful basis — usually specific consent, or employment purposes under section 7(i) if proportionate — Rule 6 safeguards such as template encryption and access logs, and no tracking of children.

  • Biometrics are ordinary personal data under the DPDP Act, but sensitivity affects SDF designation and penalties.
  • Facial recognition for customers generally needs specific consent and a non-biometric alternative.
  • Employee attendance biometrics may rely on section 7(i) only if necessary and proportionate.
  • Store encrypted templates, not raw images, and keep keys separate from template stores.
  • Section 9(3) bars tracking and behavioural monitoring of children outside Fourth Schedule exemptions.
  • A leaked biometric template cannot be reset, so breach exposure is long-lived.

Who this is for: Retailers and malls using cameras, Employers with biometric attendance, Fintech and KYC providers, Building security and access-control vendors, Computer vision startups

How does the DPDP Act treat facial recognition and biometrics?

The DPDP Act protects all digital personal data the same way. It does not create a special category for biometric, health or financial data, unlike the SPDI Rules 2011 (which listed biometric information as sensitive) and GDPR Article 9. A face image, a face embedding, a fingerprint template or an iris scan of an identifiable person is simply personal data.

That does not make biometrics low-risk under DPDP:

  • Sensitivity is a statutory factor for designating Significant Data Fiduciaries (section 10(1)) and for the Board in setting penalties.
  • Breach impact is permanent. A password can be changed; a face cannot. Rule 7 breach intimations must describe likely consequences and safety measures — hard to offer for leaked templates.
  • Security penalties are the highest in the Schedule, up to ₹250 crore per instance for failing to take reasonable security safeguards.
  • Children's rules bite. Facial tracking of children engages the section 9(3) ban on tracking and behavioural monitoring.

Separately, Aadhaar biometric authentication is governed by the Aadhaar Act and UIDAI regulations, which are outside this page; check those directly if you use Aadhaar-based face or fingerprint authentication.

The SPDI Rules' biometric protections continue until section 43A of the IT Act is omitted on 13 May 2027. See the SPDI transition.

Lawful basis by use case

Use caseLikely basisKey conditions
Face login or selfie verification in an appConsent (s.6)Specific, separate from terms; offer an alternative such as OTP
Video KYC for regulated entitiesConsent, plus sector rulesFollow the regulator's current KYC directions
Employee attendance (face or fingerprint)s.7(i) employment purposes, or consentNecessity and proportionality; consider card or app alternatives
Office access controls.7(i), consent for visitorsVisitors are not employees; give notice and an alternative
Retail store face recognition for loss preventionConsent is hard to obtain meaningfully; legal analysis neededCovert identification of shoppers is high-risk
Event or venue entryConsentDelete templates after the event
Photo-tagging in a consumer appConsentDefault off; easy withdrawal
School or crèche safety monitoringFourth Schedule Part A exemption from 9(1) and 9(3), limited to safety or educational activities9(2) detriment rule still applies

Consent must be free, specific, informed, unconditional and unambiguous. A biometric system that is a condition of service when a non-biometric alternative would work is hard to defend as "free" consent. Recording the alternative you offer is useful evidence.

For employees, section 7(i) covers processing for the purposes of employment, but continuous face tracking at desks or emotion analysis goes well beyond attendance. See HR and employee data.

Protecting biometric templates

Rule 6(1) sets the minimum safeguards. For biometric systems, apply them strictly:

  • (a) Encryption and tokens. Store templates, not raw images, wherever the system allows. Encrypt templates at rest with keys in a separate KMS or HSM. Where possible, use template-protection schemes that allow revocation, or on-device matching so templates never leave the user's device.
  • (b) Access control. Separate the matching service from the enrolment database; no human should browse templates. Restrict vendor support access with time-bound approvals.
  • (c) Logs and monitoring. Log every enrolment, match, failed match and template export. Alert on bulk template reads.
  • (d) Continuity. Back up template stores encrypted, and plan for revocation and re-enrolment if keys are compromised.
  • (e) One-year retention of logs for investigating unauthorised access.
  • (f) Vendor contracts that bar vendors from using your images to train their models without authorisation, and require deletion on instruction.
  • (g) Organisational measures — DPIA-style risk review before deployment, bias and accuracy testing, and staff training.

Minimise raw image retention. If a camera captures many people to identify a few, process non-matches transiently and discard them immediately, logging only the event.

Accuracy, bias and decisions

Facial recognition makes errors, and error rates can differ across demographic groups, lighting and camera quality. Under section 8(3), where personal data is likely to be used to make a decision affecting the Data Principal, the Data Fiduciary must ensure its completeness, accuracy and consistency. A false match that blocks someone's account, denies building entry or flags a shopper as a suspect is such a decision.

Practical controls:

  1. Test before deployment on data representative of your users and conditions; record false-match and false-non-match rates by group.
  2. Set thresholds by risk. Higher thresholds for decisions with serious effects.
  3. Human review for any adverse decision based on a match, with the image evidence available to the reviewer.
  4. Fallback paths for people who fail matching repeatedly.
  5. Grievance route. Rule 14(3) requires a published grievance period of no more than 90 days; for access-blocking errors, aim for same-day resolution.

Once a fiduciary is notified as a Significant Data Fiduciary, Rule 13(3) adds a duty of due diligence that technical measures, including algorithmic software, are not likely to pose a risk to Data Principals' rights. See algorithmic due diligence.

Children, CCTV and public spaces

Children

Section 9(3) bars tracking and behavioural monitoring of children. Face recognition that tracks children's movements or behaviour is caught unless an exemption applies. The Fourth Schedule permits:

  • Educational institutions — tracking and behavioural monitoring for educational activities or the safety of enrolled children.
  • Crèches and child day-care centres — tracking and behavioural monitoring in the interests of safety.
  • Transport providers engaged by them — location tracking during travel to and from the institution.

These exemptions lift parental consent (9(1)) and the tracking ban (9(3)) only within those conditions. The 9(2) ban on processing likely to harm a child's well-being still applies — for instance, sharing classroom face analytics with third parties.

CCTV and public spaces

Ordinary CCTV that records identifiable people is personal data processing. Adding face recognition turns passive recording into identification. Post clear signage, describe the purpose, limit retention, and restrict who can run searches. Covert identification of the public by private operators is legally and reputationally high-risk and needs careful legal analysis.

EU overlap

For systems used in the EU, Article 50 of the EU AI Act requires deployers of emotion recognition or biometric categorisation systems to inform people exposed to them.

Rights, withdrawal and erasure

Data Principals can withdraw consent at any time, with ease comparable to giving it (Rule 3(c)(i)). For biometrics:

  • Provide a one-step "delete my face/fingerprint" control where enrolment was self-service.
  • On withdrawal or end of purpose, erase templates in your systems and cause vendors to erase theirs (section 8(7)); keep only the processing logs needed for the Rule 8(3) one-year minimum, not the template itself, unless counsel advises otherwise.
  • For employees leaving the organisation, delete attendance and access templates as part of exit processing.
  • Respond to access requests (section 11) with a summary of the biometric processing and the vendors involved.
  • Include the contact person's business details in every response (Rule 9).
  • Expect correction requests too: a person whose face repeatedly fails to match, or who has been wrongly matched to someone else, is asking you to correct inaccurate processing under section 12. Re-enrolment and removal of the wrong association should be quick.
  • Track every erasure through a ticket with timestamps per system and per vendor, so you can show completion if the request later reaches the Board as a complaint.

Further reading on suhasbhairav.com

Before 13 May 2027

Compliance checklist

  1. List every biometric system

    Record capture points, templates, matching services, vendors, regions and the people affected.

    Section 8(1)

  2. Document the lawful basis per use

    Choose consent or section 7(i) for each use and record why it is necessary and proportionate.

    Sections 6, 7(i)

  3. Offer a non-biometric alternative

    Provide OTP, card or manual options so biometric consent is genuinely free.

    Section 6(1)

  4. Store encrypted templates, not images

    Encrypt templates with separately held keys and discard non-match images immediately.

    Rule 6(1)(a)

  5. Log enrolments, matches and exports

    Keep biometric access logs for at least one year and alert on bulk reads.

    Rule 6(1)(c); Rule 8(3)

  6. Test accuracy and bias before launch

    Measure error rates across groups and set thresholds and human review for adverse decisions.

    Section 8(3)

  7. Check children's exposure

    Disable face tracking of children unless a Fourth Schedule exemption applies, and review for detriment.

    Section 9; Rule 12

  8. Build template deletion

    Let users delete biometric data easily and propagate erasure to vendors.

    Section 8(7); Rule 3(c)(i)

Key dates

  1. DPDP Rules notified

  2. DPDP obligations apply to biometric systems; IT Act s.43A and SPDI protections end

Answers

Frequently asked questions

Is facial recognition legal in India under DPDP?

Facial recognition is not banned by the DPDP Act, but it is processing of personal data. It needs a lawful basis, usually specific consent for customers or employment purposes for staff if proportionate, plus Rule 6 safeguards. Tracking children is barred except under narrow Fourth Schedule exemptions.

Is biometric data sensitive personal data under DPDP?

No. The DPDP Act has no sensitive data category, so biometric data is protected as ordinary personal data. Sensitivity still matters for Significant Data Fiduciary designation and penalties. The SPDI Rules, which list biometrics as sensitive, apply until 13 May 2027.

Can employers use biometric attendance without consent?

Possibly, under section 7(i), which permits processing for employment purposes. The processing should be necessary and proportionate, with templates secured and deleted when employment ends. Offering a non-biometric alternative strengthens your position and helps where consent is the basis.

Can shops use face recognition on customers in India?

It is high-risk. Customers would generally need to give specific, free consent, which is hard to obtain meaningfully for in-store identification. Clear signage, purpose limits, short retention and legal review are essential, and covert identification of shoppers is difficult to justify.

What happens if biometric templates are leaked?

It is a personal data breach. Rule 7 requires intimation to each affected person and to the Board without delay, with a detailed report within 72 hours. Failure to take reasonable security safeguards can attract a penalty of up to ₹250 crore per instance.

Can schools use facial recognition on students?

Educational institutions are exempt from parental consent and the tracking ban only for tracking and behavioural monitoring for educational activities or student safety, under the Fourth Schedule. Processing likely to harm a child's well-being remains prohibited, and other uses need verifiable parental consent.

Legal references and sources

  • Digital Personal Data Protection Act, 2023 — Sections 6, 7(i), 9(3) (official text)
  • Digital Personal Data Protection Rules, 2025 — Rule 6 (official text)
  • Digital Personal Data Protection Rules, 2025 — Rule 12 and Fourth Schedule (official text)
  • Digital Personal Data Protection Rules, 2025 — Rule 13(3) (official text)

This guide explains the law for engineering and business teams. It is not legal advice; confirm how it applies to your organisation with qualified counsel, and check the Gazette for later notifications or amendments.

About the author

Suhas Bhairav

Suhas Bhairav

Systems architect and applied AI practitioner who designs production AI systems for Indian and global companies, with a focus on privacy-by-design under the DPDP Act and the EU AI Act.