Quick answer
Section 3(c)(ii) of the DPDP Act excludes personal data that the Data Principal herself made publicly available, or that someone published under a legal obligation in India. It is not a blanket exemption for anything on the internet. Data posted by third parties, leaked data and data behind logins are not automatically excluded, so scrapers and AI teams must check the source.
- The exclusion covers only data made public by the Data Principal herself, or published by someone legally obliged to publish it.
- Something being on the internet does not, by itself, take it outside the DPDP Act.
- Third-party posts, news reports, leaks and data behind logins need their own lawful basis.
- Once you combine public data with private data, the enriched record is likely back in scope.
- Scraping still has to comply with other laws, such as copyright, the IT Act and contracts with the website.
Who this is for: AI and ML teams building training datasets, Web scraping and data enrichment companies, Recruitment, lead generation and people-search platforms, OSINT and threat intelligence teams, Legal and compliance teams reviewing data sourcing
What does the DPDP Act say about publicly available data?
Section 3(c) lists processing to which the Act does not apply. Clause (ii) covers personal data that is made or caused to be made publicly available by:
- (A) the Data Principal to whom the data relates; or
- (B) any other person under an obligation under any law for the time being in force in India to make the data publicly available.
The Act illustrates limb (A) with an individual who, while blogging her views, has publicly made available her personal data on social media. Processing that data falls outside the Act.
Why this is narrower than it sounds
The test is who made the data public, not whether it is public. Two common misreadings cause problems:
- "It is on the open web, so it is exempt." It is exempt only if the person herself, or a legally obliged publisher, put it there.
- "It was public when we scraped it, so it is exempt forever." The Act does not say what happens if she later makes it private or deletes it. Do not assume a permanent pass. Re-check before reuse.
When the exclusion applies, the whole Act does not apply to that processing. There is no notice, consent, rights or breach duty. That makes it valuable, and also the first thing the Board would examine in a complaint.
Which public data is excluded? A source-by-source matrix
| Source | Likely excluded under s.3(c)(ii)? | Reasoning |
|---|---|---|
| A user's own public social media post or bio | Often yes | Made public by the Data Principal; confirm the account is genuinely hers |
| Her own public blog, portfolio or GitHub profile | Often yes | Self-published |
| Statutory registers that the law requires to be published | Often yes | Limb (B), a legal obligation; check the specific statute |
| Court or tribunal orders published by the court | Check | The publisher's legal obligation must be shown; sector rules may restrict reuse |
| News articles about her | Unclear to no | Published by a third party, usually without a legal obligation to publish |
| Someone else's post mentioning or tagging her | No | Not made public by her |
| Leaked or breached datasets | No | Not made public by her or under a legal obligation |
| Data behind a login or paywall, or from a closed group | No | Not publicly available in the ordinary sense |
| Data she shared with a restricted audience, such as friends only | No | Not made publicly available |
| Data inferred about her from public posts, such as political leaning or income band | Risky | The inference is new data that she did not publish |
"Often yes" still needs evidence. Store the source URL, capture time and the basis for believing the publisher was the Data Principal or a legally obliged entity.
Web scraping under DPDP: practical rules
Scraping pipelines usually mix sources. A single crawler may pick up self-published profiles, third-party mentions and stray contact details from directories. Treat the exclusion as something you prove per record or per source class, not a property of the crawler.
Design patterns that hold up
- Source allow-lists. Crawl only source types where limb (A) or (B) is defensible, such as a person's own profile page, not comment threads about her.
- Provenance fields. Store
source_url,fetched_at,publisher_type(self, legal obligation or third party) andvisibilitywith every record. - Field-level filtering. Drop personal fields you do not need. Minimisation is cheap and reduces risk if your reliance on the exclusion fails.
- Refresh and tombstone. Re-crawl periodically and delete records whose source has been removed or made private.
- Respect access controls. Do not bypass logins, rate limits or technical blocks. Apart from DPDP, unauthorised access can raise issues under the IT Act and the site's terms. Check with counsel.
If the exclusion does not apply
You need a lawful ground. Consent is impractical for most scraped data, and section 7's legitimate uses rarely fit commercial scraping. Unless another exemption applies, the realistic options are to not collect, to collect with consent, or to anonymise at ingestion.
Remember too that DPDP is only one layer. Copyright in content, database rights under contract, and platform terms are separate questions.
Training AI models on public personal data
There is no AI or machine learning training exemption in the DPDP Act or Rules. AI teams have three possible routes.
Route 1: section 3(c)(ii) public data
Training on data that Data Principals themselves made public, such as their own public posts, can fall outside the Act. Limits:
- the dataset must exclude third-party, leaked and restricted content, which large web crawls rarely achieve without filtering;
- enrichment with first-party user data brings the combined data back in scope;
- the Act does not say whether the exclusion survives later deletion by the user, so build a takedown and retraining or unlearning process.
Route 2: research exemption (section 17(2)(b) and Rule 16)
From 13 May 2027, Rule 16 says the Act does not apply to processing necessary for research, archiving or statistical purposes if it follows the Second Schedule standards. These are lawful processing, purpose limitation, data minimisation, accuracy, limited retention, security safeguards and accountability. Whether commercial model training is "research" is not settled. Do not rely on this route without legal advice. See research exemption.
Route 3: consent or no personal data
Collect with Rule 3 notice and consent, or strip and anonymise personal data before training.
Outputs matter too
A model that regurgitates a person's phone number or address is processing personal data when it produces that output. Whatever route you used for training, add output filters for identifiers and a channel for Data Principals to report and suppress outputs. See LLM training data under DPDP for the full training-data workflow.
Enrichment, lead generation and people-search platforms
Enrichment businesses take public signals, such as a professional profile, and attach them to customer records or sell profiles to clients. Three DPDP issues arise.
1. Combination brings data back in scope
When a fiduciary merges excluded public data with data it collected itself, such as a CRM email or purchase history, the merged profile is processing of personal data it holds under the Act. Notice, purpose limitation, rights and security obligations apply to the merged record.
2. Downstream clients are fiduciaries too
A client buying an enriched profile decides why and how to use it, so it is a Data Fiduciary for that processing. The client cannot rely on the vendor's claim that "it was public". It needs its own basis for its own use, such as consent before a marketing call.
3. Rights requests will arrive
Once a record is in scope, the person can ask for access under section 11, including the identities of other fiduciaries the data was shared with. She can ask for correction or erasure under section 12, and complain under section 13. Rule 14 requires you to publish how to make requests and to respond to grievances within a period of no more than 90 days.
The practical consequence is that people-search and lead-generation platforms should assume full DPDP compliance for their India operations, rather than building on the public data exclusion.
Steps for scrapers and AI teams before 13 May 2027
The Act's obligations for in-scope data apply from 13 May 2027. Datasets assembled now will still be in use then, so build the controls into collection today.
- Classify every source as self-published, legally mandated, or third-party and restricted. Only the first two can support section 3(c)(ii).
- Record provenance at ingestion and keep it with the record or training shard.
- Separate pipelines for excluded public data and in-scope data, so that merging is a deliberate, reviewed step.
- Minimise and filter: remove contact details, identity numbers and children's data unless you have a specific, documented need.
- Exclude likely children's content. Data about children triggers section 9 obligations once in scope, including the bar on tracking and behavioural monitoring.
- Build takedown handling: a public form, a suppression list, and a re-index or retrain cadence.
- Publish a contact under Rule 9 so that individuals can raise concerns. This is required anyway for any in-scope processing.
- Review with counsel if you depend on the research exemption, or on public data for a core commercial product.
Further reading on suhasbhairav.com
- BlogAI Web Scraping Agents: Adaptive Extraction vs Fixed Selectors
- BlogData Minimization vs Data Retention: Practical Guidelines
- BlogManaging LLM copyright risks in production AI systems
- LLMsKalike (ಕಲಿಕೆ): Pure Kannada LLM Trained From Scratch
- Prompts100 Best ChatGPT Prompts for Web Scraping Script Layouts
Before 13 May 2027
Compliance checklist
Tag every source by publisher type
Record whether the data was self-published, published under a legal obligation, or published by a third party.
DPDP Act s.3(c)(ii)
Store provenance with each record
Keep the source URL, fetch time and visibility so that reliance on the exclusion can be shown later.
DPDP Act s.3(c)(ii)
Block leaked and access-controlled data
Exclude breach dumps, login-gated pages and restricted-audience posts from ingestion.
DPDP Act s.3(c)(ii)
Treat merged profiles as in scope
Apply notice, rights and Rule 6 security controls once public data is combined with first-party data.
Rule 6; s.11–12
Filter children's data
Remove content likely to relate to under-18s to avoid section 9 obligations.
DPDP Act s.9
Build a takedown and suppression process
Delete or suppress records and model outputs when sources are removed or individuals object.
Rule 14
Get legal sign-off before relying on Rule 16
Confirm that your use is research, archiving or statistics and that it meets the Second Schedule standards.
Rule 16; Second Schedule
Key dates
Rules notified; Act definitions in force
Core obligations and Rule 16 research exemption apply
Answers
Frequently asked questions
Is scraping publicly available personal data legal under the DPDP Act?
Is scraping publicly available personal data legal under the DPDP Act?
Only partly exempt. Section 3(c)(ii) excludes data the Data Principal herself made public, or data published under a legal obligation. Scraped third-party posts, leaks and restricted content are not excluded and need a lawful basis. Other laws, such as copyright, the IT Act and website terms, apply separately.
Can I train an AI model on public social media data in India?
Can I train an AI model on public social media data in India?
Data that users themselves made publicly available can fall outside the DPDP Act under section 3(c)(ii). Third-party content, private posts and leaked data cannot. There is no specific AI training exemption, so filter sources, record provenance and handle takedown requests, or anonymise data before training.
Does the public data exemption apply to news articles about a person?
Does the public data exemption apply to news articles about a person?
Usually not by itself. The exclusion depends on who made the data public. A news report is published by a third party, not by the Data Principal, and normally not under a legal obligation to publish. Treat such data as in scope unless another ground or exemption applies.
What if a user deletes a post I already collected?
What if a user deletes a post I already collected?
The DPDP Act does not say expressly whether the section 3(c)(ii) exclusion survives once the Data Principal withdraws data from public view. The safer practice is to re-check sources periodically, delete or suppress records whose source was removed, and plan model retraining or output suppression.
Is data in government registers exempt under DPDP?
Is data in government registers exempt under DPDP?
Data published by a person under an obligation under Indian law to make it public falls within section 3(c)(ii)(B). Check that the specific register is published under such a legal obligation. Some statutes also restrict how published data may be reused, independently of the DPDP Act.
Is enriched B2B lead data exempt because it came from public profiles?
Is enriched B2B lead data exempt because it came from public profiles?
Not once you combine it with your own data or use it to contact people. The merged profile is personal data you process under the Act, and clients buying it act as Data Fiduciaries. Notice, lawful basis, rights and security obligations apply from 13 May 2027.
Legal references and sources
- Digital Personal Data Protection Act, 2023 — Section 3(c)(ii) (official text)
- Digital Personal Data Protection Act, 2023 — Section 17(2)(b) (official text)
- Digital Personal Data Protection Rules, 2025 — Rule 16 and Second Schedule (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
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.
Related guides
- AI & personal dataTraining LLMs on Personal Data Under the DPDP ActTraining or fine-tuning a model on personal data is processing under the DPDP Act. Here is how lawful basis, public data, research, erasure and memorisation work.
- DPDP Act 2023Who Does the DPDP Act Apply To? Territorial Scope ExplainedSection 3 of the DPDP Act decides who is in scope: anyone processing digital personal data in India, and offshore businesses offering goods or services to people in India.
- DPDP Act 2023What Counts as Personal Data Under the DPDP ActHow the DPDP Act defines personal data and digital personal data, how identifiability works for IDs, logs and AI data, and why India has no separate sensitive data category.
- DPDP Rules 2025DPDP Exemption for Research, Archiving and Statistics (Rule 16)Rule 16 switches off the DPDP Act for processing necessary for research, archiving or statistics, but only if you follow the Second Schedule standards. Here is what that means in practice.
- DPDP Act 2023DPDP Consent Requirements: Free, Specific, Informed and UnambiguousWhat valid consent means under section 6 of the DPDP Act and Rule 3: the five qualities, clear affirmative action, necessity, withdrawal parity and proof of consent.
- DPDP Act 2023DPDP Section 17 Exemptions ExplainedSection 17 switches off parts of the DPDP Act for legal claims, courts, crime, offshore outsourcing, mergers, loan defaults, notified State bodies, research and notified startups.