India has unveiled a new system designed to enable companies and researchers to train artificial intelligence (AI) algorithms using sensitive data from hospitals, banks, and government agencies without compromising the security of that information. The Data Empowerment and Protection Architecture (DEPA) for AI Training was launched by the Ministry of Electronics and Information Technology (MeitY) and developed by the volunteer-led technology think tank, iSPIRT.

The DEPA framework aims to keep data within its original institutions while allowing AI models to be trained on it. This approach contrasts with traditional methods where data is often transferred to researchers, raising concerns about privacy and security. A senior government official highlighted that this model prioritises data protection while facilitating AI development.

Although DEPA is a new application of existing principles, it builds on a framework introduced in 2020 by NITI Aayog, which allowed citizens to share their data securely with consent. The latest iteration extends this concept to include the training of AI models on sensitive data without the need for data transfer.

Consultations and Legal Framework Development

Consultations regarding the implementation of DEPA for AI are set to begin in the coming weeks, as stakeholders work towards establishing a legal framework. This process is expected to take several years. The launch of DEPA coincides with India’s preparation to enforce its first data privacy law, the Digital Personal Data Protection (DPDP) Act, 2023, which is anticipated to take effect in the coming months.

During the launch, industry representatives expressed optimism about the framework’s potential to facilitate the development of domain-specific AI applications. Jibu Elias, India lead for responsible computing at the Mozilla Foundation, noted that while the framework could be crucial, building trust and validating the quality of sensitive data will be significant challenges.

Under DEPA, the data remains with its original owner, while a secure computing environment, referred to as a “sealed room,” operates under strict digital contracts. This setup ensures that no unauthorised access to the data occurs, including by the operators of the system. Developers can submit AI models and receive only the trained models or specific decisions, such as those related to loan applications.

Initial Pilots and Industry Reactions

Initial pilot projects have already been demonstrated, showcasing the framework’s capabilities. One pilot involved a bank that tested the system to identify eligible customers for credit cards, processing over 2 million requests in a two-week period. Another pilot focused on a health-tech startup training a tumour-detection model using MRI scans, with the data remaining inaccessible to the startup itself.

Key figures at the launch included Ajay K Sood, principal scientific adviser to the government, and representatives from various organisations, including HDFC Bank and Emergent AI. Stakeholders highlighted the importance of the framework, particularly as hospitals and banks are often reluctant to share data due to financial and legal concerns.

Currently, the management of DEPA is expected to be overseen by Samyog, a new industry group supported by iSPIRT. The group will register participants, standardise contracts, and address disputes. However, details regarding payment structures for data usage and liability in case of data leaks remain unspecified.

The framework is not legally binding at present, and the timeline for establishing binding regulations is still uncertain. Future developments will likely depend on the implementation of the DPDP Act and ongoing discussions regarding AI legislation in India.

Long or Short, get news the way you like. No ads. No redirections. Download Newspin and Stay Alert, The CSR Journal Mobile app, for fast, crisp, clean updates!

App Store – https://apps.apple.com/in/app/newspin/id6746449540

Google Play Store – https://play.google.com/store/apps/details?id=com.inventifweb.newspin&pcampaignid=web_share