Vendor Risk / DPA

Vendor Data Processing Agreement

Negotiate a data-processing agreement covering instructions, security, subprocessors, breaches, rights support, retention, deletion, audit, location and exit.

A security brochure does not define what a SaaS vendor may do with customer data.

Quick View

Decision

Convert the data map and risk assessment into enforceable processor obligations.

First action

Map vendor data flow.

Core evidence

Executed DPA.

Main warning

Vendor owns broad reuse rights.

Why It Matters

The final Digital Personal Data Protection Rules, 2025 were notified on 14 November 2025 with phased commencement. As of 25 June 2026, organisations should distinguish provisions already commenced from operational duties scheduled for later dates, while continuing to comply with the IT Act, CERT-In directions and sectoral rules already in force.

The Act states that a Data Fiduciary may engage a processor under a valid contract and remains responsible for processing carried out on its behalf.

The agreement should match the actual service architecture, support flows, subprocessors, hosting regions and retention behaviour.

Control Framework

AreaWhat to establishOperating rule
ScopeData, subjects and processing purpose.Attach schedule.
SecurityAccess, encryption, logging and testing.Use measurable duties.
IncidentNotification, cooperation and evidence.Set rapid timeline.
ExitReturn, deletion and transition.Verify completion.

Action Checklist

  1. Map vendor data flow.
  2. Review subprocessor list.
  3. Set breach escalation.
  4. Define rights support.
  5. Agree retention and deletion.
  6. Test termination export.

Practical Example

A payroll SaaS contract promises ‘industry-standard security’ but gives no incident timeline, subprocessor list or deletion commitment.

Evidence to Keep

  • Executed DPA.
  • Security schedule.
  • Subprocessor register.
  • Audit reports.
  • Incident and request tickets.
  • Exit and deletion certificate.

Warning Signs

  • Vendor owns broad reuse rights.
  • Unlimited retention.
  • No breach deadline.
  • Unapproved subprocessors.
  • No export or deletion process.

Detailed Review

Privacy governance should connect the personal data, individual, purpose, collection point, system, owner, recipient, access role, retention trigger and incident dependency. A policy that cannot be traced to this chain is difficult to operate.

Create a dated legal matrix rather than one status label. Record the DPDP provision, commencement date, present readiness action, current IT or sectoral obligation and the evidence owner.

Design controls in the product and system. A written rule cannot stop an SDK from firing, a shared folder from exposing payroll, or a vendor from retaining deleted users unless technology and operations enforce it.

Evidence should be generated during normal work: versioned notices, event logs, access approvals, request tickets, deletion reports, vendor registers, incident chronologies and management decisions.

Use proportionate identity and security checks. Excess verification creates more personal data, while weak verification can expose another person’s records or permit account takeover.

Vendor risk should be tiered by data sensitivity, volume, access privilege, business criticality and replacement difficulty.

Contract clauses should be tested against actual product capability, especially export, deletion, logging, subprocessor and incident functions.

Control Test

Test the control using a real user journey from collection to deletion. Capture the notice shown, data stored, vendors called, employees with access, retention period and response if the user withdraws or complains.

Run a negative scenario: the vendor is breached, the user is a child, the employee exits, the phone is stolen, the data was inaccurate or the regulator asks for proof. Record which control fails.

Check that system records and public wording agree. Product forms, privacy notice, CRM fields, SDK behaviour, vendor contracts and support scripts should describe the same processing.

Assign a named owner and internal deadline for every gap. A risk register without funded action and closure evidence becomes an archive of known failures.

Retain the rejected alternatives and decision basis. This is especially important where the law is in phased commencement or a proportionate technical method is selected.

Escalation Route

Start with the system owner, privacy or security owner and the documented data flow. Preserve records before making changes, and separate current statutory reporting from future DPDP readiness.

For a breach, financial fraud, rights dispute, children’s-data issue or regulated-sector event, involve qualified legal, cyber, forensic and sector specialists and use the applicable official reporting or grievance channel.

Management Review

Management should record the risk owner, affected data population, financial or operational impact, current legal duty, future DPDP milestone and funded remediation date. A privacy register without accountable closure is only a list of known gaps.

The control should be tested with evidence rather than self-certification. Use screen recordings, exported logs, access reports, deletion output, vendor responses, tabletop minutes or complaint acknowledgements to prove that the workflow operates as designed.

Where several laws apply, maintain one incident or request chronology but separate each legal trigger and deadline. CERT-In, RBI, UIDAI, IRDAI, police, contractual and DPDP processes should not be collapsed into one generic notification decision.

Common Questions

Does a vendor’s compliance certificate replace a DPA?

No.

Who is responsible to the user?

The Data Fiduciary remains responsible for processing on its behalf.

Should subprocessors be controlled?

Yes, through transparency and contractual controls.

What is the most important exit term?

Usable data return followed by verified deletion.

Official Sources

Use the latest commencement notifications, final Rules, CERT-In directions and sectoral regulator material. Applicability depends on dates, roles, systems, users and facts.

Disclaimer: This article is educational and does not provide personal legal, privacy, cyber-forensic, banking, insurance, employment or regulatory advice. Obtain qualified advice before implementing or reporting a material issue.