Guide · Data Protection
The DPDP Act and Rules: what they mean for IT and AI projects
India's data protection law is moving into force in phases. What IT and AI teams should build into systems now, before the main obligations apply in 2027.
In short: India's Digital Personal Data Protection Act, 2023 is being brought into force in phases under the DPDP Rules, 2025, which were notified in November 2025. The main operational obligations, including notice, security safeguards, breach reporting and retention, apply 18 months after notification, in May 2027. IT and AI teams should use the time to map personal data, build consent and notice into systems, tighten security and logging, and update vendor contracts. This article is general information, not legal advice.
The phased timeline
MeitY notified the DPDP Rules on 13 November 2025 with a three-phase rollout. Provisions setting up the Data Protection Board took effect immediately; consent-manager registration follows after twelve months; and most operational obligations apply after eighteen months, in May 2027. The earlier SPDI Rules under the IT Act remain in force until then. Organisations should confirm the exact dates for each provision with their legal advisers.
What changes for system design
- Data mapping: know what personal data each system holds, where it came from, why it is processed and who it is shared with.
- Notice and consent: clear notices at the point of collection, consent recorded with time and purpose, and an easy way to withdraw it.
- Purpose and retention: keep data only as long as the purpose needs, with deletion built into systems rather than done by hand.
- Security safeguards: encryption, access control, logging and monitoring proportionate to the data.
- Breach readiness: the ability to detect, assess and report breaches to the Board and affected people.
- Rights requests: processes to give people access to, correct or erase their data.
- Children's data: verifiable parental consent where children's data is processed.
What it means for AI projects
- Check the purpose for which data was collected before using it to train or ground AI models.
- Minimise personal data in prompts, retrieval indexes and logs.
- Apply access controls to retrieval, so an assistant cannot reveal records a user may not see.
- Record which data sources each AI system uses, so rights requests and deletions can reach them.
Contracts with vendors and processors
Organisations that decide why and how data is processed (data fiduciaries) remain responsible for processors acting on their behalf. Contracts with IT, cloud, AI and support vendors should cover security measures, breach notification to the client, sub-processors, deletion at the end of the contract and audit rights. Devices returned at refresh should be wiped to policy, with certificates.
Nuclear Edge handles personal data under the DPDP Act in India and, abroad, under PIPEDA and the GDPR. See our commitments, practical AI governance and private LLM deployment.
Frequently asked questions
When do the DPDP Act obligations apply?
The DPDP Rules were notified on 13 November 2025 with phased commencement. Most operational obligations, such as notice, security safeguards, breach reporting and retention, apply 18 months later, in May 2027. Confirm exact dates with legal advisers.
Does the DPDP Act apply to AI training data?
Where AI systems process digital personal data, the Act's principles apply, including purpose limitation, security and data principals' rights. Teams should check the lawful basis and purpose before using personal data in AI.
What should a vendor contract include under the DPDP Act?
Security obligations, breach notification to the data fiduciary, controls on sub-processors, deletion or return of data at the end of the contract, and audit rights.
Is this legal advice?
No. This article is general information for IT and AI teams. Organisations should take advice from qualified lawyers on their own obligations.