Qualysec

MeitY Unveils DEPA for AI Training: What It Means for AI Security and Data Governance in India

Tue Nov 03 2026
MeitY Unveils DEPA for AI Training What It Means for AI Security & Data Governance in India

Data Empowerment and Protection Architecture Handbook for AI Training has been released by the Ministry of Electronics and Information Technology (MeitY) and the iSPIRT Foundation. The MeitY Secretary, S. Krishnan launched the framework in Bengaluru to solve the “Data Deadlock.”  It is an issue related to regulatory and security barriers that prevent startups, researchers, and businesses from safely using sensitive, sector-specific data to train AI models.

DEPA changes how data is processed. Instead of moving sensitive data to the AI model, the model is brought to the data.

The DEPA framework comprises four pillars, which are Data Exchange Platforms, Electronic Contracts, Secure Data Environment (SDEs), and Differential Privacy. The DEPA framework aims to make critical data sets available for AI training by protecting sensitive data. The first deployment will focus on the healthcare and BFSI sectors.

The Cybersecurity Imperative: Why SDEs & AI Pipelines Demand Audit-Grade Security

DEPA offers a structure for securing data sharing; however, putting DEPA into use will also create security issues for engineering and security teams. These issues must be solved to secure sensitive data and AI infrastructure.

1. Securing Clean Rooms and Confidential Environments

AI models trained in isolated Secure Data Environments (SDEs) still require strong security measures. These include checking cloud configurations, controlling who can access the data, and continuously monitoring for threats to prevent sensitive information from being exposed.

2. API and Interoperability Vulnerabilities

DEPA uses APIs to connect data providers, consent managers, and AI training environments. If these APIs are not properly secured, attackers may exploit weak endpoints or broken object-level authorization (BOLA) to access data or systems they are not authorized to use.

3. AI Model Privacy and Integrity Risks

Privacy-Enhancing Technologies (PETs), such as differential privacy, help protect sensitive information during AI training. However, AI models can still face threats such as data poisoning, membership inference, and model inversion attacks if their training and deployment processes are not properly secured.

How Qualysec Helps You Build DEPA-Ready Data Pipelines

As organizations like the healthcare industry, fintech, and enterprise tech get ready to comply with DPDP and DEPA standards, security validation becomes a necessity.

Qualysec assists enterprises in securing their data exchanges and AI applications by:

  • AI/ML Penetration Testing: Evaluation of model security against vulnerabilities such as data theft, poisoning, and privacy bypass.
  • API Security Assessment: Examination of consent frameworks and interfaces for transferring data to prevent unauthorized data exposure.
  • Cloud and Infrastructure Assessment: Securing the Secure Data Environment (SDE) and Confidential Computing Infrastructure from lateral movement.

Make sure that your security is not only proven but also integrated into your AI and compliance workflows.

Talk to Qualysec’s Cybersecurity Experts Today.

#DEPA AI security

Get a Quote

Let's work together to secure your business!

Please fill out the form to let us know about your cybersecurity needs and our professionals will reach out shortly to discuss your unique needs.

Total No. Of Vulnerabilities

0+

Total No. Of Vulnerabilities

Years in Business

0+

Years in Business

Assessment Completed

0+

Assessment Completed

Trusted Clients

0+

Trusted Clients

Countries Served

0+

Countries Served

Subscribe to Newsletter