Sector guidesUpdated 22 November 202611 min read

DPDP Compliance Checklist for SaaS & AI Startups (2026–27)

A 12-point DPDP checklist and month-by-month plan for SaaS and AI startups to be ready when the core DPDP Rules apply on 13 May 2027.

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

Core rules apply
13 May 2027
Already in force
Board rules (13 Nov 2025); Consent Manager registration (13 Nov 2026)
Breach report to Board
Detailed report within 72 hours (Rule 7(2)(b))
Log retention
Minimum one year from processing (Rule 8(3))
Grievance response
Published period, max 90 days (Rule 14(3))
Highest penalty
Up to ₹250 crore (security safeguards)
Startup exemption
Only if notified under s. 17(3)

Quick answer

SaaS and AI startups have until 13 May 2027, when DPDP Rules 3 and 5–16 take effect, to be compliant. The essentials: map personal data and your fiduciary or processor role, fix notices and consent, sign processor contracts, implement Rule 6 safeguards, set up 72-hour breach reporting, configure one-year log retention then erasure, handle children's data, and publish rights and grievance channels.

  • Core obligations — notice, security, breach, retention, children, rights — apply from 13 May 2027, about six months from now.
  • B2B SaaS is often a Data Processor for customer data but a Data Fiduciary for its own users, leads and employees.
  • Every personal data breach is reportable; there is no materiality threshold, and the detailed Board report is due within 72 hours.
  • Rule 8(3) requires personal data, traffic data and processing logs to be retained for a minimum of one year, then erased.
  • There is no general startup exemption today; section 17(3) only allows the Government to notify one.

Who this is for: SaaS founders and CTOs, AI and GenAI startups, B2B platform companies, Startup legal and compliance leads, Engineering managers

What applies to a SaaS or AI startup, and when?

The DPDP Rules, 2025 (G.S.R. 846(E), 13 November 2025) commence in three phases under Rule 1:

DateWhat is in forceStartup impact
13 Nov 2025Rules 1, 2, 17–21: Board constitution and procedureNone operationally
13 Nov 2026Rule 4: Consent Manager registrationOptional — you may integrate with registered Consent Managers
13 May 2027Rules 3, 5–16, 22, 23: notice, security, breach, retention, contact, children, SDF, rights, transfers, appeals, government requestsEverything below

The Act's own operative sections follow the same calendar: the remaining sections, including penalties, come into force on 13 May 2027.

Until then, section 43A of the IT Act and the SPDI Rules, 2011 still apply; section 44(2) of the DPDP Act omits section 43A from 13 May 2027.

There is no blanket startup exemption. Section 17(3) lets the Central Government exempt classes of fiduciaries, including startups, from certain provisions by notification — plan as if no exemption will arrive.

The Act applies to digital personal data processed in India, and to processing outside India connected with offering goods or services to people in India (section 3). See territorial scope.

Step one: fiduciary or processor?

Your obligations depend on your role for each data set.

  • Data Fiduciary — you decide purpose and means. For a SaaS company: your own sign-ups, account admins, website leads, marketing lists, employees, and any consumer-facing product.
  • Data Processor — you process on a customer's behalf. For B2B SaaS: the end-user data your customers upload or generate in your product.

Most B2B SaaS companies are both. The fiduciary obligations attach to your direct relationships; the processor side is governed mainly by the contracts your customers sign with you, because section 8(1) makes them responsible for what you do on their behalf and section 8(2) requires a valid contract.

As a processor, expect enterprise customers to demand: Rule 6 safeguards, prompt breach notice so they can meet Rule 7, one-year retention of data and logs under Rule 8(3) (the Rule's cloud-provider illustration describes exactly this), erasure on instruction, and sub-processor transparency.

AI startups have a third question: are you using customer data to train your own models? If so, for that purpose you are acting as a fiduciary and need a lawful basis of your own. See Data Fiduciary vs Data Processor.

Where you rely on consent, section 6(1) requires it to be free, specific, informed, unconditional and unambiguous, with clear affirmative action, and limited to necessary data. Withdrawal must be as easy as giving consent.

Rule 3 sets the notice content:

  • Presented and understandable independently of other information.
  • Clear and plain language.
  • An itemised description of the personal data.
  • The specified purposes and a specific description of the goods, services or uses enabled.
  • A link to your website or app, and other means, to withdraw consent, exercise rights and complain to the Board.

Section 5(2) also requires a notice, as soon as reasonably practicable, to people whose consent was obtained before the Act commenced. For a startup with an existing user base, that means a re-notice campaign before or around 13 May 2027.

Not everything needs consent. Section 7(a) covers data voluntarily provided for a specified purpose where the person has not indicated non-consent; section 7(i) covers employment purposes.

Typical startup fixes: separate marketing and product-analytics consent from account creation; replace "we may share with partners" with named categories; add a settings toggle that mirrors the opt-in. See the privacy notice template.

Security safeguards and breach readiness

Rule 6(1) sets minimum safeguards that apply to your systems and your processors':

  • (a) encryption, obfuscation, masking or virtual tokens;
  • (b) access control to computer resources;
  • (c) logs, monitoring and review to detect, investigate and remediate unauthorised access;
  • (d) continuity measures such as backups;
  • (e) one-year retention of those logs and personal data;
  • (f) security clauses in processor contracts;
  • (g) technical and organisational measures to ensure the safeguards work.

Rule 7 breach notice has no materiality threshold. On becoming aware of any personal data breach you must:

  • Tell each affected Data Principal without delay: what happened, likely consequences, mitigation, what she can do, and a contact.
  • Tell the Board without delay with a description, then within 72 hours a detailed report covering circumstances, mitigation, findings on who caused it, remediation and the user notifications sent.

CERT-In's April 2022 Directions separately require reporting specified incidents within 6 hours of noticing them.

For a startup, this means an incident runbook with named owners, a pre-drafted user notice, and a tabletop drill before May 2027. See the breach response playbook.

AI-specific controls for startups

If your product uses LLMs, embeddings or agents, add these to the baseline:

  • Model gateway. All LLM calls pass through one service that masks identifiers, selects providers and logs consistently. See masking and tokenisation for AI pipelines.
  • Provider terms. Confirm each model API's retention and no-training settings; record them in your vendor register.
  • Vector DB hygiene. Tag chunks with tenant and subject IDs so you can erase on request; schedule compaction after deletes. See erasure in LLMs and vector databases.
  • Trace logging. Keep metadata and data-access references for a year; mask content; erase after. See LLM logging under the one-year rule.
  • Training data lineage. Know which customer records feed which fine-tunes or eval sets, and on what basis.
  • Agent tool scopes. Agents that call CRMs or databases need least-privilege credentials and logged tool calls.
  • Children. If minors could use your product, you need age assurance, parental consent under Rule 10 and no behavioural profiling of under-18s.

Selling to Europe as well? The EU AI Act's Article 50 transparency obligations apply from 2 August 2026 and GDPR applies to EU users. See selling AI SaaS to Europe.

Phased plan: 22 November 2026 to 13 May 2027

About 25 weeks remain. A realistic plan for a team of 20–200 people:

Phase 1 — Discover (late Nov to mid Dec 2026)

  • Appoint an owner; set a weekly 30-minute review.
  • Build the data inventory: systems, data categories, purposes, role, processors, regions. See data mapping.
  • List every vendor and AI provider that touches personal data.

Phase 2 — Design (mid Dec 2026 to end Jan 2027)

  • Draft Rule 3 notices and consent flows; decide legitimate-use reliance per purpose.
  • Define retention classes, including Rule 8(3) one-year logs and erasure jobs.
  • Write the breach runbook and the rights-request workflow.

Phase 3 — Build (Feb to mid Apr 2027)

  • Ship consent and withdrawal UX, rights-request intake, and Rule 9 contact publication.
  • Implement Rule 6 gaps: encryption, access reviews, logging, backups.
  • Sign or update processor contracts; respond to customer DPA requests.
  • Configure log retention and erasure lifecycle policies.
  • Build age assurance and parental consent if children may use the product.

Phase 4 — Prove (mid Apr to 13 May 2027)

  • Run a breach tabletop and an end-to-end erasure drill.
  • Send re-notices to existing users under section 5(2).
  • Publish the grievance timeline (max 90 days) and request channels.
  • Collect evidence in one folder: inventory, notices, contracts, drills, policies.

Mistakes startups make

  • Assuming B2B means DPDP does not apply. Your admins, leads and employees are your Data Principals, and customers will contractually push processor duties onto you.
  • Waiting for a startup exemption. Section 17(3) permits one; none should be assumed.
  • Treating breach notice like GDPR. GDPR Article 33 notification turns on risk; DPDP Rule 7 covers every personal data breach.
  • Deleting logs too fast — or never. Rule 8(3) is a one-year minimum, then erasure.
  • Shadow AI. Engineers or support staff pasting customer data into personal AI accounts outside your contracts.
  • No DPO, no contact. Ordinary fiduciaries do not need a DPO, but Rule 9 requires a published contact person who can answer processing questions.
  • Ignoring Consent Managers. Optional for you, but since 13 November 2026 registered Consent Managers can exist; be ready to accept consents through them if customers ask.

See also the broader compliance roadmap to May 2027 and the DPDP audit checklist.

Further reading on suhasbhairav.com

Before 13 May 2027

Compliance checklist

  1. Build a personal data inventory

    Record systems, data categories, purposes, your role, processors and regions for every data flow, including AI pipelines.

    DPDP Act s. 8(1)

  2. Classify your role per data set

    Mark where you are Data Fiduciary and where you are Data Processor, and note any use of customer data for your own model training.

    DPDP Act s. 2

  3. Rewrite notices to Rule 3 standard

    Standalone, plain-language, itemised data list, specific purposes, and links to withdraw, exercise rights and complain to the Board.

    Rule 3

  4. Unbundle consent and make withdrawal easy

    Separate marketing and analytics consent from account creation and add a withdrawal toggle as easy as opt-in.

    DPDP Act s. 6; Rule 3(c)(i)

  5. Re-notice existing users

    Send a DPDP notice to users whose consent predates commencement, as soon as reasonably practicable.

    DPDP Act s. 5(2)

  6. Close Rule 6 security gaps

    Implement encryption or masking, access control, monitoring, backups and organisational measures across your stack.

    Rule 6(1)(a)–(g)

  7. Sign processor contracts with security clauses

    Ensure every vendor, including LLM APIs and observability tools, is under a valid contract with safeguard and retention terms.

    DPDP Act s. 8(2); Rule 6(1)(f)

  8. Prepare 72-hour breach reporting

    Write a runbook covering user notices without delay, Board notice and the detailed 72-hour report, plus CERT-In's 6-hour reporting.

    Rule 7

  9. Configure one-year log retention and erasure

    Keep personal data, traffic data and processing logs for at least one year from processing, then erase automatically.

    Rule 8(3)

  10. Publish a contact person

    Show the DPO or contact person's business details on your site or app and in every rights-request response.

    Rule 9

  11. Handle children's data correctly

    Add age assurance, verifiable parental consent and no tracking or targeted ads for under-18 users if minors may use the product.

    DPDP Act s. 9; Rule 10

  12. Publish rights and grievance channels

    Explain how to request access, correction, erasure and nomination, which identifiers are needed, and a grievance period of at most 90 days.

    Rule 14

Key dates

  1. DPDP Rules notified; Board rules (1, 2, 17–21) in force

  2. Rule 4 Consent Manager registration in force

  3. Suggested: Phase 1 discovery and data inventory complete

  4. Suggested: Phase 2 notices, retention and runbooks designed

  5. Suggested: Phase 3 build complete; contracts signed

  6. Rules 3, 5–16, 22 and 23 and remaining Act sections in force

Answers

Frequently asked questions

When does DPDP compliance become mandatory for startups?

The core obligations apply from 13 May 2027, when Rules 3, 5–16, 22 and 23 and the remaining sections of the Act come into force. That covers notices, security safeguards, breach notification, retention, children's data, rights and grievances. Board rules have applied since 13 November 2025 and Consent Manager registration since 13 November 2026.

Is there a DPDP exemption for startups?

Not automatically. Section 17(3) of the DPDP Act allows the Central Government to exempt certain Data Fiduciaries or classes, including startups, from specified provisions by notification, having regard to the volume and nature of data processed. Unless such a notification covers you, plan for full compliance by 13 May 2027.

Does a B2B SaaS company need to comply with DPDP?

Yes. You are a Data Fiduciary for your own users, admins, leads and employees, and usually a Data Processor for your customers' end-user data. As a processor, your customers remain responsible under section 8(1) and will require contract terms covering safeguards, breach notice, retention and erasure.

Does my startup need a Data Protection Officer?

Only Significant Data Fiduciaries must appoint a DPO under section 10, and no SDF notification is assumed. Every other fiduciary must publish the business contact details of a person who can answer questions about processing, under Rule 9, and include them in every response to a rights request.

How is DPDP breach notification different from GDPR?

GDPR Article 33 requires notification to a supervisory authority within 72 hours where there is risk to individuals. DPDP Rule 7 has no risk threshold: every personal data breach must be reported to affected individuals without delay and to the Board, with a detailed report within 72 hours of becoming aware.

What does DPDP compliance cost a small SaaS startup?

The main costs are engineering time for consent, rights, logging and erasure features, legal time for notices and contracts, and any tooling such as consent management or log storage for one year. Costs scale with the number of systems and vendors, which is why the data inventory comes first.

Can I use customer data to train my AI models under DPDP?

Only with a lawful basis for that purpose. If you process customer data as a processor, training your own models on it is a new purpose for which you act as a fiduciary. Get contractual permission from customers and confirm the underlying consent or legitimate use covers it, or train on masked or synthetic data.

Legal references and sources

  • Digital Personal Data Protection Rules, 2025 — Rule 1 (commencement) (official text)
  • Digital Personal Data Protection Rules, 2025 — Rules 3, 6, 7, 8, 9, 10, 14 (official text)
  • Digital Personal Data Protection Act, 2023 — Sections 5–9, 11–14, 16, 17(3) (official text)
  • Digital Personal Data Protection Act, 2023 — Schedule (penalties) (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.