3 hrs ago
NTA Uses Sovereign AI for NEET-UG Translation After Paper Leak
The National Testing Agency used its own computers to help translate questions for a medical entrance exam retest.
It did not send the material to cloud services.
Officials said this helped protect private and sensitive information.
The first NEET-UG exam was cancelled after a paper leak.
Investigators said people with access to the translated paper were involved.
The agency is now working toward safer, internet-free systems with human checks.
Speakers at a panel also discussed building more data centres in India.
They said more centres are needed, but these facilities use electricity and water.
They also said smaller AI systems can help with public services such as health, farming and education.
The National Testing Agency used sovereign AI models on its own GPUs to translate material for the NEET-UG retest.
NTA avoided cloud systems, citing data privacy and information-security concerns.
The original NEET-UG examination was cancelled after investigators linked the paper leak to insiders with access to translated papers.
India’s data-centre capacity is far below its data-generation needs, according to NTA Director General Abhishek Singh.
Panelists supported expanding AI infrastructure while acknowledging data centres’ energy and water demands.
- Who
- The National Testing Agency, led in the discussion by Director General Abhishek Singh, used sovereign AI models for translation.
- What
- NTA used on-premise GPUs rather than cloud systems to translate material for the NEET-UG retest and discussed broader AI infrastructure needs.
- Where
- The remarks were made at the India Habitat Centre in New Delhi.
- When
- The original NEET-UG examination was held on 3 May and cancelled on 12 May; Singh’s remarks were made at the Charcha 2026 panel on Tuesday.
- Why
- NTA cited data privacy and information security, while the translation process had been scrutinized after the original examination’s paper leak.
AI and Data-Centre Expansion
Privacy and Environmental Caution
Building more data centres
AI and Data-Centre Expansion
Abhishek Singh said India’s data generation far exceeds its data-centre capacity and argued that ecological concerns should not currently deter expansion. Shalini Kapoor said renewable energy, including solar and wind power, could run data centres.
Privacy and Environmental Caution
Singh acknowledged that data centres consume substantial electricity and water. The Economic Survey 2025-26 described them as a double-edged sword because of those resource demands.
Cloud systems versus sovereign infrastructure
AI and Data-Centre Expansion
Using sovereign models on on-premise GPUs keeps sensitive examination material under the agency’s control and avoids reliance on cloud services.
Privacy and Environmental Caution
Cloud systems were not used because Singh said data privacy and information security were important; the articles do not present a separate argument in favor of cloud use for this examination.
Key facts
- Agency
- National Testing Agency
- Exam
- NEET-UG
- Original examination
- Held on 3 May for about 22 lakh candidates
- Cancellation
- The examination was cancelled on 12 May after a paper leak
- Translation system
- Sovereign AI models running on NTA’s own, on-premise GPUs
- IndiaAI GPU deployment
- More than 38,000 GPU units deployed, compared with an initial target of 10,000
- Data-centre gap
- India produces about 20% of the world’s data but hosts roughly 3% of global data-centre capacity
Quotes
Abhishek Singh
Director General of the National Testing Agency
“The entire translation… was done with our sovereign AI models… we use an on-prem version of it on GPUs… without using the cloud version”
theprint.in
“I don’t think ecological concerns should at this stage deter us”
theprint.in









