Within the most recent 72 hours assembly electionsElection Commission of India (ECI) released detailed statistical reports and index cards ECINet CapabilitiesIts digital election platform formally launches in its full commercial version in January 2026 Special Intensive Review (SIR) for that west bengal electionsDespite all relevant information being readily available within ECINet, the ECI neither disclosed the status of approximately 34 lakh appeals, including seven lakh annulment appeals, pending before 19 Supreme Court-appointed tribunals, nor issued a comparable transparency report.
Reportedly, a tribunal headed by the former Chief Justice of the Calcutta High Court disposed of 1,777 appeals, allowing all 1,717 civil appeals for inclusion and rejecting all 60 ECI appeals for deletion. Similar large-scale reforms may be in store for the other 18 jurisdictions. In contrast, the ECI had reportedly covered only about 1,607 voters before the poll.
Such selective disclosures and inconsistent transparency raise serious questions about the functioning of the ECI, neutrality and equally the suppression of important information that could potentially influence electoral outcomes.
Despite this, there is a lack of comparable transparency and status disclosure for SIR. Report of massive discrepancies And disenfranchisement raises serious questions about selective disclosure and institutional double standards within the ECI. As a constitutional authority, the ECI enjoys the highest institutional dignity and trust. Equally, it must remain accountable, transparent and open to objective scrutiny, especially when allegations of arbitrariness, bias and massive exclusion of genuine voters emerge.
In this backdrop, an independent AI-enabled inspection layer integrated with ECINet can continuously assess the electoral roll revision processes, including neutrality, consistency and procedural arbitrariness. The proposed AI watchdog framework is easy to implement, a basic operational model can be achieved within a few months and is capable of continuous enhancement thereafter.
Failures in SIR 2.0
SIR 2.0 exposed unprecedented chaos induced by ad-hoc, constantly changing and subjective SOPs (Standard Operating Procedures), which reportedly excluded millions of genuine voters from the electoral rolls and, in many cases, deprived them of candidature rights. What began as an exercise to improve electoral accuracy by removing ASDD (absent, transferred, duplicate, dead) entries and adding new voters instead, resulted in widespread uncertainty, repeated verifications, lengthy appeals, and allegations of arbitrariness, discrimination and bias.
The exercise relied heavily on the inaccurate, incomplete and non-searchable legacy SIR 2002-04 database. Instead of correcting flaws at the source, the burden of proof was placed on voters, forcing genuine citizens to establish their eligibility again and again despite long voting histories and valid documents. This process was marked by uneven application of logical discrepancy criteria across regions and voter groups, resulting in non-uniform outcomes for similarly situated voters. Minor mismatches in name, age or family details often lead to exclusion, while opaque decision-making and the absence of rational orders have led to allegations of arbitrariness and algorithmic bias.
The results were most worrying in West Bengal, where only 1,600 inclusion appeals and only six deletion cases out of approximately 3.4 million pending appeals were disposed of before the vote, even though the success rate of inclusion appeals was reportedly over 99%. Election officials and prospective candidates were reportedly among those excluded. Notably, one such excluded person, who was later cleared for inclusion, was elected as an MLA.
In an unprecedented situation, the victory margin in 49 assembly constituencies was reportedly less than the number of voters awaiting disposal of inclusion appeals. The top court said relief for many may come only in future elections and post-poll scrutiny may be necessary in constituencies where the margin of victory is less than the scale of anomalies and pending appeals, raising serious concerns over electoral integrity and the possibility of post-poll chaos. This situation reflects not only administrative failure but also a deep crisis of credibility in the electoral roll revision process.
These developments exposed deep structural weaknesses in voter list management. They stood in contrast to the Election Commission’s repeated commitment to “ensure free, fair, transparent, accessible and peaceful elections” and its assurance that “no genuine voter is disenfranchised.”
More importantly, this happened despite ECINet reportedly being able to handle three crore hits per minute and maintaining detailed operational data for every voter and transaction. Yet neutrality, consistency and accountability remained largely dependent on opaque manual processes, administrative discretion and ex post facto corrections. Therefore, the SIR 2.0 experience underlined the urgent need for a sustainable, technology-driven oversight mechanism capable of monitoring processes, detecting anomalies, assessing institutional neutrality, and identifying discriminatory patterns in real time.
AI Inspection for ECINet
As AI is increasingly powering governance and large-scale public systems, election management also requires intelligent, continuous auditable oversight. Incorporating an AI-enabled watchdog within ECINet provides a practical path to creating a neutrality-aware voter list management system capable of protecting democratic participation and public trust.
Directly integrated with ECINet, the proposed AI layer will act as a continuous monitoring and analysis engine. It will monitor usage of the system, track decision patterns, analyze voter-official interactions and generate real-time indicators of neutrality, consistency, efficiency and citizen satisfaction at the booth, constituency, district and state levels. Unlike ex post reviews, it can continuously audit electoral roll revision processes using transactional and procedural data already available within ECINet, allowing irregularities to be detected early, before they escalate into large-scale disenfranchisement or administrative crises.
The system can automatically flag anomalies and discriminatory patterns, including abnormal increase in deletions, inconsistent application of SOPs, trend of repeated rejections linked to specific authorities, excessive complaint delays, sudden policy changes, bias arising from logical anomaly filters, disproportionate exclusions due to minor spelling or family-data mismatches, and concentrated deletions in specific areas, castes or communities. It can also compare results across regions to identify differential treatment with similarly placed voters and enforce consistency in institutional communications by tracking announcements, circulars, deadlines, SOP amendments, and field instructions.
Furthermore, ongoing analysis of bottlenecks, software glitches, validation failures, complaint trends, and operational inefficiencies can support evidence-based refinement of SOPs, replacing ad-hoc administrative responses with measurable corrective action. A basic operational version of such an AI watchdog could be implemented within a few months and then continuously scaled up.
way forward
An AI-enabled watchdog integrated with ECINet can continuously monitor electoral operations, assess institutional neutrality, detect anomalies and discriminatory patterns, and flag anomalies or changes in eligibility criteria.
It can standardize declarations, timelines and procedural updates, reducing confusion and non-uniform implementation across regions.
With ECINet already fully operational, AI-powered inspection can make SIR processes more transparent, neutral, accountable and citizen-centric. Importantly, such a system would strengthen – not replace – constitutional authority through transparent audit trails, fairness metrics, evidence-based oversight and measurable accountability, while reducing arbitrariness, ambiguity and public distrust.
(Rajeev Kumar is a former Professor of Computer Science at IIT Kharagpur, IIT Kanpur, BITS Pilani and JNU and a former scientist of DRDO and DST)
