← All news and guides
Data protection

What a DPDP + AI Technical Audit Actually Delivers: Sample Scope, Evidence and Fix List

A DPDP technical audit, from the inside: a sample scope with AI tools included, the method, each deliverable, a sample gap register and what it does not cover.

Mono Team ·

Printed pages, a closed notebook and an open laptop on a wooden desk in a Bengaluru office at dusk, city lights blurred through the window.

A DPDP technical audit shows you, one system at a time, where your handling of personal data falls short of the DPDP Act and Rules, and what to fix first. Ours ends with five things: a data map, a gap register tied to the section or rule behind each gap, the evidence we checked, a prioritised fix list with owners and effort, and a list of questions for legal review. It does not give you a legal opinion or a certificate.

We call this work a DPDP gap assessment. This post shows what one looks like from the inside: an illustrative scope, the method, each deliverable, and a sample of the gap register.

Technical audit, gap assessment, readiness assessment: one job

People use these names for the same piece of work. None of them is defined in the Act. The only audit the Act names is the data audit that a Significant Data Fiduciary must have done by an independent data auditor (Section 10). That is a different job, for a small group of companies the Government designates. We compare the two in who can do a DPDP audit in India.

For everyone else, a technical audit is how you find out what to build before the core duties start on 13 May 2027. It looks at what your systems actually do with personal data, not at what your policies say they do.

A sample scope

Scope is agreed in writing before we start. The table below is illustrative: a made-up consumer app company with a web app, a mobile app and a small data team. It is not a real client. Yours will have different systems.

SystemIn scope?What we look at
Production databaseIn scope?: InWhat we look at: Personal data fields, encryption, who can query it, how long rows are kept
Mobile app and analytics SDKsIn scope?: InWhat we look at: What each SDK collects and sends, and whether consent is asked first
CRM and email or SMS vendorIn scope?: InWhat we look at: Consent flags, how withdrawals reach the vendor, what the vendor receives
Support deskIn scope?: InWhat we look at: Ticket text, attachments, retention, and which AI features are switched on
Data warehouse and dashboardsIn scope?: InWhat we look at: Copies of personal data, who can see them, and whether deletion reaches them
Cloud logs and backupsIn scope?: InWhat we look at: What is logged, how long logs are kept, backup encryption and restore tests
HRMS and payrollIn scope?: InWhat we look at: Employee and applicant data, and which vendors hold it
AI toolsIn scope?: InWhat we look at: See the list below
Paper records that are never digitisedIn scope?: OutWhat we look at: The Act covers digital personal data only
Penetration testing and code security reviewIn scope?: OutWhat we look at: A separate security engagement
Vendors' own systemsIn scope?: OutWhat we look at: We review what each vendor receives and what its contract says, not its internal controls

AI tools are in scope by default

The Act does not mention AI. It applies to the personal data you put into AI tools, the same as any other vendor. AI tools are often where personal data goes that nobody listed. We check:

  • Company-approved AI assistants used by staff, and what goes into prompts and uploads.
  • AI coding assistants with access to repositories, logs or database copies.
  • LLM API calls inside your product, such as a support bot or a feature that summarises customer records.
  • AI features inside SaaS tools you already use, such as a CRM or support desk, which may be on by default.
  • Personal AI accounts used for work. We ask about this in interviews. We don't monitor staff.

For each one we record what personal data goes in, under which contract and settings, how long the vendor keeps it, and whether anything is masked before it leaves your systems.

The method

Two people's hands over a meeting table with a notebook, a pen and a closed laptop, in soft evening light.

We work with read-only access. Nothing in your systems changes during the assessment. We use four ways of finding facts:

  • Interviews. Short conversations with the people who run each system: engineering, product, support, marketing, HR, and whoever buys software. They tell us what people believe happens.
  • Data discovery. We read schemas, sample exports you approve, and storage listings to find personal data fields, including the ones nobody listed: free-text notes, attachments, and logs that capture request bodies.
  • Config review. The settings that decide what happens to data: encryption, access roles, log retention, backup policies, deletion jobs, SDK settings and the scopes of vendor integrations. They tell us what actually happens.
  • Data-flow mapping. We follow one person's data from the moment it is collected to every copy, every vendor and every point where it should be deleted.

Then we compare each finding against the Act and Rules. Where an interview and a config disagree, the config wins and the difference becomes a finding. The steps are set out on the gap assessment page: kickoff, discovery interviews, system review, gap analysis, then a readout with your team.

The deliverables

A stack of printed reports bound with clips beside a cup of chai on a dark desk, a window behind them out of focus.

1. The data map

An inventory of every system that holds personal data: what it holds, why, who owns it, where it flows, which vendors receive it, whether it leaves India, and how long it is kept. It comes with data-flow diagrams. Everything else depends on it. You cannot answer a rights request, delete on schedule or scope a breach without knowing where the data is.

2. The gap register

One row per gap. Each row has the finding, the section of the Act or rule it relates to, a severity, and a pointer to the evidence. Severity reflects how likely the gap is to harm people and which duty it touches. A sample is below.

3. The evidence list

For each finding, what we looked at: a schema, a setting, a vendor's terms, an interview note. For each duty, the record you would need to show that it works, such as a consent log extract, retention settings, dated access reviews or a tested breach runbook. If your customers send you DPDP questionnaires, the same records go into a vendor evidence pack.

4. The prioritised fix list

Each gap turned into work: what to change, who owns it, a relative effort, what it depends on, and the order to do it in. The gap assessment page calls this the costed implementation plan. You can hand it to your own team, or to us for DPDP implementation.

Some questions are legal, not technical, and we don't answer them. We write them down with the facts a lawyer needs, and our partner law firm reviews them. Typical items:

  • Which basis fits each use of personal data: consent, or one of the legitimate uses in Section 7.
  • Whether a vendor acts only on your behalf as a Data Processor, or uses the data for its own purposes.
  • What the processor contracts need to say, including security terms (Section 8(2), Rule 6(1)(f)).
  • Whether an exemption in Section 17 applies to any of your processing.
  • Where another law requires you to keep data that you would otherwise delete.

You also get a one-page summary your board can read.

A sample gap register

These rows are illustrative. They come from the made-up company above, not from a real client, and your register will differ. They show the format and the kind of finding a technical audit produces.

#FindingAct / RulesSeverityFixOwnerEffort
G1Finding: Sign-up uses one checkbox for the service, marketing and analytics. No record of which notice version the user saw.Act / Rules: Sections 5 and 6(1), 6(10); Rule 3Severity: HighFix: Consent per purpose, versioned notices, and a consent logOwner: Product and backendEffort: Medium
G2Finding: Unsubscribing in the app does not reach the email vendor. Marketing continues.Act / Rules: Sections 6(4) and 6(6)Severity: HighFix: Send withdrawal events to every system and vendor that uses the consentOwner: BackendEffort: Small
G3Finding: Support agents paste ticket text into personal AI chat accounts. No contract with the AI vendor.Act / Rules: Sections 8(2) and 8(5); Rule 6Severity: HighFix: Approve one AI tool on company terms, mask identifiers, block personal accounts. Contract terms to legal review.Owner: IT and support leadEffort: Medium
G4Finding: Application and database access logs are kept for 30 days.Act / Rules: Rule 6(1)(c) and 6(1)(e)Severity: MediumFix: Keep access logs for one year in a store the logged systems cannot editOwner: PlatformEffort: Small
G5Finding: Closed accounts are deleted in the app database but stay in the warehouse and backups indefinitely.Act / Rules: Section 8(7); Rule 8(3)Severity: MediumFix: Retention period per system and scheduled deletion jobs, with the one-year minimum. Conflicting retention laws to legal review.Owner: Data teamEffort: Large
G6Finding: Staging is refreshed with an unmasked copy of production every week.Act / Rules: Section 8(5); Rule 6(1)(a)Severity: MediumFix: Mask or synthesise personal data in stagingOwner: PlatformEffort: Medium
G7Finding: The incident runbook covers CERT-In's 6-hour report but not the Data Protection Board or affected people.Act / Rules: Section 8(6); Rule 7Severity: HighFix: One runbook for both, with templates and a drillOwner: Security leadEffort: Small

Effort is relative: Small, Medium and Large. The fix list breaks each item into tasks with an estimate. Items G3 and G5 also appear in the legal review list, because part of each fix depends on a legal answer.

How the fix list gets its order

Severity is one input, not the only one. We order the work like this:

  • Records that must already be running on 13 May 2027. A consent log only proves what it recorded, and Section 6(10) puts the burden of proof on you. One-year log retention is the same. Start these first.
  • Security gaps with a high severity. The highest penalty in the Schedule, up to ₹250 crore, is for failing to take reasonable security safeguards to prevent a breach. Our Rule 6 checklist lists the minimum safeguards.
  • Work that crosses many systems. Deletion and rights requests touch every copy of the data, so they take longest.
  • Text that waits on legal review. Notices and contract terms are drafted once the lawyers have answered the flagged questions.

Our compliance calendar shows how this fits into the months left before 13 May 2027.

What a DPDP technical audit does not deliver

  • A legal opinion. We tell you what your systems do. Whether that meets the law in your case is for your lawyers. Our partner law firm gives the written legal opinion.
  • A certificate. The Act and Rules do not set up a DPDP compliance certificate. What you can show a customer or the Board is evidence.
  • The independent data audit for a Significant Data Fiduciary. If the Government designates you, Section 10 requires an independent data auditor. In our view, a team that helps you build the fixes should not also audit them.
  • A penetration test or code review. Those are separate security engagements.
  • Changes to your systems. The assessment is read-only. Fixing is a separate step.

Before you book one

If you are not sure DPDP applies to you, take the 2-minute DPDP self-check first. If it does, the assessment goes faster when you have three things ready:

  • A rough list of your systems and the vendors that receive personal data, including AI tools.
  • One engineer who can give read-only access and answer questions.
  • An hour or so from each system owner for interviews.

New to the law itself? Start with the DPDP Act explained. See everything we do for DPDP compliance.

This article is general information, not legal advice. Our engineers do the technical work. Our partner law firm reviews it and gives the written legal opinion.

Sources: Digital Personal Data Protection Act, 2023, Sections 5 to 8, 10 and 17 and the Schedule. Digital Personal Data Protection Rules, 2025, Rules 3, 6, 7, 8 and 13 (G.S.R. 846(E), Gazette of India, 13 November 2025). CERT-In Directions, 28 April 2022. Last reviewed 18 November 2026.

FAQ

What is a DPDP technical audit?

It is a review of the systems that hold personal data against the DPDP Act and Rules. It finds where personal data lives, what each system actually does with it, and where that falls short. It ends with a data map, a gap register, an evidence list, a prioritised fix list and questions for legal review.

What is included in a DPDP gap assessment?

A data map of every system and vendor that holds personal data, including AI tools. A gap register that ties each gap to a section of the Act or a rule, with a severity. The evidence behind each finding, a fix list with owners and effort, the items that need legal review, and a one-page board summary.

Does a DPDP technical audit include AI tools?

Ours does, by default. The Act applies to personal data put into AI tools the same as any other vendor. We check approved AI assistants, coding assistants, LLM calls in your product, AI features inside SaaS tools, and whether staff use personal accounts for work.

Is a DPDP technical audit a legal opinion or a certificate?

No to both. It tells you what your systems do and where they fall short. Legal questions go to a law firm, and the Act and Rules do not set up a DPDP compliance certificate. What you get is evidence you can act on and show to others.

Do you change our systems during the assessment?

No. We work with read-only access and nothing in your systems changes. Fixing the gaps is a separate step, which your team or ours can do.

See what your systems do with personal data.