Key points
- India’s MeitY proposes DEPA to enable local AI training without exposing raw data to external entities.
- The scheme uses hardware-based secure environments where data remains encrypted during processing.
- Differential privacy techniques are described as systematically applied to prevent individual data leakage.
India’s Ministry of Electronics and Information Technology has proposed a new regulatory framework for artificial intelligence training that keeps sensitive information within secure domestic boundaries. According to a report by The Register, the initiative aims to foster local AI development while preventing offshore model makers from gaining unrestricted access to national datasets.reportedThe plan keeps sensitive data locked away while letting AI programs come to it for training.The framework envisions model mobility into Secure Data Environments, applying differential privacy and immutable audit logs.
Evidence: reported
India floats AI plan under which ‘The model travels to the data, rather than the data travelling to the model’aifleet-register
The claim
The proposal, termed Data Empowerment and Protection Architecture for AI Training, reverses the conventional data flow. As described by Professor Ajay Kumar Sood, Principal Scientific Adviser to the Government of India, the core principle is that the model travels to the data, rather than the data travelling to the model. This approach seeks to ensure that owners of data retain control while addressing national issues through locally trained systems.
The response
The Register reports that the scheme envisions Training Data Providers, such as hospitals, banks, and government agencies, sharing anonymized and encrypted data with a Data Exchange Platform.reported Training Data Consumers then access this stored information to train their models. The platform operates within hardware-based Secure Data Environments, ensuring data remains encrypted and isolated even during active processing, while enforcing usage rules set by providers.
Evidence: reported
India floats AI plan under which ‘The model travels to the data, rather than the data travelling to the model’aifleet-register
The bottom line for teams
MeitY’s framework document states that differential privacy techniques are systematically applied to prevent models from learning or revealing individual-specific information. All activities inside these secure environments are immutably logged for audit purposes, ensuring the process is both secure and transparent. However, it remains unclear how much work remains to make DEPA a reality, despite India already operating several data exchanges.
What the paper shows
The material outlines a structural shift in AI training logistics rather than presenting empirical performance results. It details a proposed architecture where secure data environments isolate encrypted datasets from external model consumers. The stated limitation is the current implementation status, with The Register noting uncertainty regarding the remaining work required to realize the DEPA framework fully.
Why it matters
- Model evaluators must assess how hardware-based secure environments impact the latency and cost of training large models.
- The systematic application of differential privacy introduces a measurable constraint on model fidelity that requires rigorous testing against standard benchmarks.
- Immutable logging of all processing activities creates a new audit trail requirement for compliance in AI development workflows.
- The shift from data mobility to model mobility challenges existing infrastructure assumptions about data pipeline efficiency.
Topics: AI regulation, data privacy, India, model training, secure environments
Updated
Then / now / next
Offshore model-makers typically accessed global datasets without strict local control mechanisms.
India proposes DEPA to keep data encrypted and isolated within secure domestic environments.
It remains unclear how much work remains to make the DEPA framework a operational reality.
Story timeline
- Then Offshore model-makers typically accessed global datasets without strict local control mechanisms.
- 02 Nov 2026 India floats AI plan under which ‘The model travels to the data, rather than the data travelling to the model’
- Now India proposes DEPA to keep data encrypted and isolated within secure domestic environments.
- Next It remains unclear how much work remains to make the DEPA framework a operational reality.
Ask the article
Local keyword search of this article; no generated answers.Did this article give you the evidence you needed?
Your choice stays on this device. No public tally.Questions readers ask
What is DEPA for AI Training?
It is a proposed Indian government framework where AI models access encrypted data within secure environments, rather than moving data to the model.
Who provides the training data in this scheme?
Training Data Providers, including hospitals, banks, and government agencies, share anonymized and encrypted data via a Data Exchange Platform.
How does the framework protect individual privacy?
It applies differential privacy techniques systematically to prevent models from learning or revealing individual-specific information.
Connected stories
Reading receipt
- India’s MeitY proposes DEPA to enable local AI training without exposing raw data to external entities.
- The scheme uses hardware-based secure environments where data remains encrypted during processing.
- Differential privacy techniques are described as systematically applied to prevent individual data leakage.