Nilekani at Bloomberg Forum: AI Question Banks Are India's Answer to Paper Leaks

The Argument India's Exam Reformers Have Been Avoiding

On October 13, 2026, Nandan Nilekani took the stage at the Bloomberg New Economy Forum and said something many in India's examination establishment had been reluctant to acknowledge: the country's persistent exam paper leak problem is not primarily a transport security failure. It is an architectural one.

Nilekani, who heads the government-appointed task force on examination reform constituted in July 2026, argued that fixed papers set weeks in advance give bad actors both the motive and the window to obtain them. The answer, he said, is to eliminate the static paper entirely—replacing it with large, secure question banks from which papers are assembled at the last minute, potentially with AI assistance to maintain difficulty calibration and syllabus coverage.

The statement arrived five months after NEET-UG 2026 was cancelled following a paper leak that the Rajasthan SOG traced back to a "guess paper" matching roughly 120 of 410 actual exam questions. It arrived three months after the CBSE on-screen marking controversy triggered the removal of the Board's chairman and secretary. India's examination credibility had taken two body blows in a single season, and the political patience for incremental fixes had run out.

Why Static Papers Break

The leak pipeline for a printed examination paper is long. A paper set two to three months before the exam date passes through question paper vetting, printing, encryption, physical dispatch, storage at nodal centres, bank vault custody, and last-mile delivery to examination venues—often crossing multiple states and hundreds of handling points. Each handoff is an opportunity.

Nilekani's core argument is that this chain cannot be made leak-proof through tighter seals and better locks. The economic incentive to obtain a paper—coaching institutes pay crores for advance access, and the downstream fees they can charge dwarf the value of any security investment—will always outpace the cost of a breach.

A question bank model changes the incentive structure. If no paper exists until minutes before the exam, there is nothing to steal. The bank itself becomes the valuable asset, but securing a digital repository behind access controls and audit trails is categorically easier than securing a physical print run distributed across thousands of locations.

The AI Calibration Problem

The technical challenge Nilekani acknowledged is maintaining consistency across dynamically assembled papers. When a traditional paper is set, subject experts spend weeks calibrating difficulty, ensuring coverage across all syllabus units, and peer-reviewing items for errors. An AI system assembling papers on demand must replicate that process reliably.

Several large-scale examination bodies internationally have developed item response theory (IRT) based systems that can assemble papers within defined difficulty parameters from pre-calibrated item banks. The National Board of Medical Examiners in the United States uses such systems for USMLE. India's own NEET 2.0 architects are reportedly studying similar models.

The question bank approach requires upfront investment: items must be pre-validated, piloted, and calibrated for difficulty before they enter the live bank. But once a sufficiently large bank exists, the cost of generating each subsequent examination falls dramatically while security improves.

150,000 Venues: The Infrastructure Prerequisite

Nilekani did not present the question bank model as a standalone solution. Secure digital delivery of dynamically assembled papers requires computer-based testing (CBT) infrastructure. He noted that India has approximately 150,000 viable test venue locations, though many would require upgrades to deliver high-stakes CBT reliably.

That number encompasses a range of institutions: National Skill Development Corporation-affiliated testing centres, engineering college computer laboratories, polytechnic institutes, and designated government examination centres. Getting all of them to a standard where candidates cannot be disadvantaged by connectivity or hardware variation is a multi-year infrastructure commitment.

The task force is expected to recommend a tiered rollout: high-stakes national examinations (NEET, JEE, CUET) moving first, followed by state-level entrance tests, followed eventually by university-level semester examinations. The latter category—semester examinations—is where on-screen evaluation of handwritten answer sheets will remain central for the foreseeable future, since handwritten answers are not going away even as delivery moves digital.

What This Means for Answer Sheet Evaluation

The Nilekani framework addresses the front-end of the examination pipeline: how questions are set and delivered. It does not address the back-end: how answers are evaluated. That is where on-screen marking (OSM) comes in, and where the CBSE controversy of 2026 demonstrated the costs of under-prepared implementation.

Even in a fully CBT-delivered world, essay-type subjective questions, open-ended problem solving, and descriptive answers will continue to require human evaluation. Universities running semester examinations will continue to scan and evaluate physical answer sheets for years after entrance tests go online. The quality of that evaluation infrastructure—scanner resolution, evaluator training, double valuation protocols, moderation workflows—determines whether the back-end is as secure and accurate as a well-implemented front-end.

The Nilekani task force's work, and the minister's endorsement of it, signals that the government is ready to invest in examination security architecture holistically. Institutions that have already built robust digital evaluation infrastructure are better positioned to align with those standards quickly.

What Universities Should Watch For

The task force's recommendations from September 13, 2026 include a mandate for standardised technology specifications for OSM platforms, requiring all examining bodies to procure systems against a published minimum specification. That specification is expected to feed into UGC circulars and potentially NAAC assessment criteria in the next review cycle.

Universities that have not yet audited their answer sheet scanning quality, evaluator authentication protocols, and tamper detection mechanisms against the emerging national standards have a narrowing window to do so. The policy architecture being built around the Nilekani recommendations will make digital evaluation quality a compliance requirement, not just a quality aspiration.

The Bloomberg forum conversation on October 13 was not a technical announcement. It was a signal about direction. India is building toward an examination ecosystem where fixed papers and physical evaluation are the exception rather than the rule. Universities that understand that direction early will spend the next two years building infrastructure. Those that do not will spend it responding to mandates.

Related Reading