Short answer: India's Digital Personal Data Protection Act, 2023 (DPDP Act) does not ban ChatGPT or any other AI tool. It regulates how organisations process personal data. So the real question is not "may we use ChatGPT?" but "may this data go into this tool, under these terms and settings?" This article gives you the checks to run before your team pastes anything in.
This is a practical guide, not legal advice. Your legal and compliance advisers should confirm how the law applies to your organisation.
Where the DPDP Act stands today
The Ministry of Electronics and Information Technology notified the Digital Personal Data Protection Rules, 2025 on 13 November 2025, operationalising the Act (HSA Advocates analysis). Implementation is phased over about 18 months. Law firm summaries describe three sets of dates: provisions effective from November 2025 (including the Data Protection Board), from November 2026 (including registration of Consent Managers), and from May 2027, when most obligations on organisations apply (AMS Shardul; AZB & Partners).
In other words, as of late 2026 the clock is running. Organisations have months, not years, to understand where personal data flows, including into AI tools.
Why AI tools fall inside the DPDP picture
Under the Act, the organisation that decides why and how personal data is processed is the "data fiduciary". A few of its core duties are directly relevant to AI tools:
- A lawful basis. Personal data may be processed for a lawful purpose with the person's consent or for certain specified legitimate uses. Using a customer's data in a new AI workflow should fit the purpose it was collected for.
- Processors act under contract. A data fiduciary may engage a data processor only under a valid contract. An AI vendor that processes your customers' personal data on your behalf is, in practice, acting as a processor.
- Reasonable security safeguards. The fiduciary must protect personal data, including data handled by its processors, with reasonable safeguards to prevent breaches.
- Erasure when the purpose is served. Data should not be kept longer than needed. That includes copies sitting in an AI tool's chat history or file storage.
- Breach reporting. Personal data breaches must be reported to the Data Protection Board and to affected individuals, in the manner the Rules prescribe.
An employee pasting a customer's loan file into a personal chatbot account can touch every one of these at once, without anyone in the organisation knowing.
Seven checks before your team uses an AI tool
- Which data is going in? Classify it: public, internal, confidential, personal data, sensitive personal data. Many uses need no personal data at all.
- Which account and plan? Consumer accounts and business plans differ sharply. For example, OpenAI states that by default it does not use data from its business offerings, such as ChatGPT Enterprise and the API platform, to train its models (OpenAI business data page). Personal accounts used for work fall outside your organisation's control.
- Is there a contract? Check whether a data processing agreement is in place, and what it says about sub-processors, security and deletion.
- Where is the data stored and processed? Storage location and processing location can differ. Some enterprise services now let eligible customers store data at rest in India; check what the option covers in your plan (OpenAI data residency announcement).
- How long is it kept? Review retention settings for chats, uploaded files and logs, and align them with your own retention rules.
- Who reviews the output? If AI output feeds a decision about a person, such as a loan, a claim or a hiring decision, a trained person should review it.
- Can you show your work? Keep a record of approved tools, the data categories allowed in each, and the reasons. If a regulator or customer asks, this is your answer.
What a sensible AI policy says
A good AI acceptable-use policy fits on a few pages. It names the approved tools and plans; lists data that must never go into them (for example customer identity documents, health information, client-privileged material); explains what needs approval; names who to ask; and, crucially, points people to a safe alternative for the work they genuinely need to do. Policies without alternatives get ignored.
If you need help writing one, our AI Governance & DPDP Readiness service produces exactly this, together with an inventory of the AI tools already in use.
When the data is too sensitive for a public tool
Some work involves data your organisation should keep entirely under its own control: KYC packs, patient records, privileged legal files. For these, the alternative is a private system that runs where you decide. A private knowledge assistant or private document AI gives people the same convenience without the data leaving your environment. Read more in Sovereign AI vs Public AI.
A practical next step
Start with visibility. List the AI tools in use across teams (including browser extensions and AI features inside existing software), note what data goes into each, and check the plan and terms. Most organisations find at least one surprise. Then write the policy, approve the tools, and give people a safe option for sensitive work.
Frequently asked questions
Does the DPDP Act ban ChatGPT?
No. The Act regulates how personal data is processed, not which tools are used. Whether a tool is appropriate depends on the data, the plan, the contract and the settings.
When do DPDP obligations apply?
The Rules were notified in November 2025 with a phased rollout over about 18 months. Most obligations on organisations apply from May 2027, according to law firm summaries of the notified timelines.
Is ChatGPT Enterprise DPDP compliant?
No tool is compliant by itself; compliance is about how your organisation processes personal data. Enterprise plans offer protections such as no training on business data by default and data residency options, which your legal team can assess against your obligations.
What should go into an AI acceptable-use policy?
Approved tools and plans, data categories that must never be entered, what needs approval, who to ask, and a safe alternative for sensitive work.
Is this legal advice?
No. This article is general guidance. Please consult your legal and compliance advisers about your specific obligations.
Sources
Written by the Novatra team
Novatra is a sovereign AI company based in Thane, India. This article is general information, not legal advice. About us