93% Turnout, 9 Million Couldn't Vote: How an Algorithm Quietly Dismantled India's Democracy
Summary
In India's 2026 West Bengal state assembly election, the Election Commission of India deployed an AI-based "Special Intensive Revision" (SIR) process that removed 9.1 million voters — 11.88% of the total electorate — from the rolls before a single ballot was cast. Among those deleted, Muslims made up 34% of all purged names despite comprising only 27% of the state's population, and in Nandigram constituency, 95.5% of deleted voters were Muslim in a district where Muslims represent just 25% of residents. Of 3.4 million objections filed, fewer than 2,000 were processed before election day, yet 98% of those reviewed were ruled "improperly deleted" — a statistical indictment of the algorithm's core premise. The BJP won West Bengal's assembly for the first time in history, securing 207 of 293 seats, but in 49 constituencies the number of deleted voters exceeded the winner's margin of victory, raising fundamental questions about electoral legitimacy. Concurrently, Freedom House docked India 14 points since 2005 and V-Dem classified it an "electoral autocracy" ranked 105th of 179 nations — together marking what may be the most thoroughly documented case of algorithmic disenfranchisement in the history of electoral democracy.
Key Points
The Scale and Structural Bias of the 9-Million Voter Purge
The sheer scale of what India's Election Commission executed in West Bengal before the 2026 election is worth sitting with for a moment. Of 76.6 million registered voters, 9,102,577 — nearly 11.88% of the entire electorate — were removed from the rolls through the Special Intensive Revision (SIR) process. That alone would be alarming in any democracy. What makes this structural rather than administrative is the distribution of who got deleted: Muslims, who make up 27% of West Bengal's population, accounted for 34% of all deletions. In Nandigram constituency, where Muslims represent just 25% of residents, a stunning 95.5% of the 2,826 deleted voters were Muslim. In Habanibpur, Muslims at 20% of the local population made up 40% of purged names. These aren't data points that scatter randomly — they form a consistent, directional pattern pointing toward something other than neutral administrative cleanup.
Zooming out to the national picture makes the numbers even harder to dismiss as coincidence or bureaucratic sloppiness. Across 12 states and union territories, SIR Phases 1 and 2 removed 51.8 million voters — roughly 10.2% of the national electorate. Uttar Pradesh alone saw 20.4 million names deleted, representing 13.21% of its voter base. The SABAR Institute's constituency-level analysis documented the same structural skew at every geographic scale: Muslim voters were consistently over-represented among the purged relative to their share of the local population. I'd argue this data doesn't describe bureaucratic error — it describes a system operating exactly as its design logic intended, regardless of whether that intention was ever committed to a planning document someone might someday subpoena.
AI's Cultural Blind Spot: How the Algorithm Targets Muslim Names
The ECI's "logical inconsistency" software works by flagging voters whose records trigger anomalies across several criteria: spelling mismatches between the 2002 and 2025 voter lists, voter-to-ancestor ratios exceeding six per household, and age gaps between voters and their parents falling outside preset ranges. On paper, this sounds like sensible data hygiene — the kind of routine maintenance that keeps bloated voter rolls accurate. In practice, it is a system designed with the cultural literacy of a majority-Hindu naming tradition applied to a multilingual, multi-script, pluralistic country where those assumptions simply don't hold.
Here is the specific mechanism producing the Muslim over-purge. Muslim names in West Bengal routinely transition between Urdu, Bengali, and English transliterations across different official documents — a single voter might appear as "Nabijan" on one government record and "Nabirul" on another, or "Sabera" in 2025 and "Chabera" in the 2002 rolls. The algorithm has no concept of transliteration drift. It reads these as two different people and flags the more recent entry for deletion. Hindu names, rooted in Sanskrit with far more stable romanization conventions across decades of official record-keeping, are far less vulnerable to this type of mismatch. The MIT Media Lab's foundational Gender Shades study (Buolamwini and Gebru, 2018) established the underlying principle: AI trained predominantly on majority-group data systematically misclassifies minority-group characteristics as anomalies. India's voter-purge algorithm is that principle's most politically consequential application to date.
The Supreme Court of India itself criticized these criteria as failing to "reflect the realities of India" — specifically noting that early marriage, late childbirth, and multilingual naming customs make age-gap and name-mismatch tests structurally unreliable across large swaths of the population. And yet the algorithm ran anyway, at scale, with no pre-deployment audit and no transparency about how it was making its decisions. The fact that 62% of deleted voters were women — many of them Muslim women whose names changed at marriage and were thus classified as entirely different individuals — only adds another layer to what is clearly a compounding intersectional bias problem.
The 0.05% Objection Rate — Due Process in Name Only
Here is the number that, for me, settles the question of whether this was administrative error or deliberate structural exclusion: 0.05%. Of 3.4 million objections filed by voters who discovered their names had been removed, fewer than 2,000 — less than one in two thousand — were processed before election day. The objections were not heard. They were acknowledged, timestamped, and shelved. That alone would be a democratic scandal by any standard. What makes it worse is the outcome of the objections that were actually reviewed: approximately 98% were ruled "improperly deleted," meaning the algorithm had wrongly flagged legitimate citizens in the overwhelming majority of cases. Extrapolate that 98% finding across the 3.4 million unprocessed objections and the implication is impossible to escape — the vast majority of the 9.1 million people removed from West Bengal's rolls were legally eligible voters.
Two testimonies from on-the-ground reporting capture what this looks like at the human scale. Muhammad Ali, a 65-year-old retired army veteran, told Jacobin: "I served in the Indian Army for this country. And now they tell me my name is not on the voter list." Nawabjan Ali, a daily-wage laborer, asked: "I work for daily wages. How many times can I leave work and go to the government office?" The objection process — with its paper forms, government office queues, and multi-day timelines — was structurally inaccessible to precisely the people most likely to have been wrongly deleted: urban poor, migrant workers, and daily-wage earners who cannot absorb the cost of bureaucratic engagement. The Association for Democratic Reforms filed suit in the Supreme Court, arguing the SIR violates Section 21(3) of the Representation of the People Act 1950 and a 1995 Supreme Court precedent that explicitly prohibits reversing the burden of proof onto voters.
Institutional Capture: How the Election Commission Lost Its Independence
To understand why an algorithm this biased could run unchecked through an election affecting 100 million people, you have to look at what happened to the institution meant to stop it. In 2023, the Modi government passed the Chief Election Commissioner and Other Election Commissioners (Appointment, Conditions of Service and Term of Office) Act, which transferred appointment authority for Election Commission members from a Supreme Court-led panel to an executive-led panel chaired by the Prime Minister himself. In plain terms: the governing party now controls who runs the elections.
This matters for the algorithm story because institutional independence is the check that would catch algorithmic bias before it becomes electoral outcome. When that check is removed, the algorithm's errors don't get corrected — they get embedded as policy. Gilles Verniers, an India politics specialist at Sciences Po, stated the dynamic precisely in a Democracy Now! interview: "The institution that is supposed to be the impartial arbiter intervenes, interferes in the electoral process, vitiates it, and creates an electoral advantage by picking Hindu voters against Muslim voters." Jawhar Sircar, a former Chief Electoral Officer of West Bengal, was less diplomatic: "The SIR is a blatant attack on democracy. The ECI has lost all credibility." Democratic Erosion, the Brown University project tracking global backsliding, classifies this as "executive aggrandizement" — the systematic absorption of nominally independent institutions into the ruling party's operational machinery. The SIR didn't run in a vacuum; it ran in a country where 115 of 121 politicians under Enforcement Directorate investigation happen to be opposition members, where the judiciary has, per BTI 2026, "repeatedly yielded to pressure from the Hindu nationalist government," and where the ECI itself now answers to the executive it is supposed to monitor.
Election Results Where Deleted Voters Outnumber the Margin of Victory
The most direct challenge to the legitimacy of the BJP's historic West Bengal victory comes from AltNews and Scroll.in's seat-by-seat data analysis, and the numbers are stark. In 49 constituencies, the number of "under adjudication" deleted voters — those whose objections were still pending when the polls closed — exceeded the winner's margin of victory. Of those 49 seats, the BJP won 26 and TMC won 21. Rajarhat New Town: BJP margin 316 votes, deleted voters 24,132. Jangipur: BJP margin 10,542, deleted voters 36,581. Scroll.in's broader analysis found that in 105 of the BJP's winning constituencies — fully half its total wins — the aggregate SIR deletions exceeded the margin of victory. These are not marginal or ambiguous results; they are constituencies where the outcome was structurally indeterminate because the precondition of universal suffrage had been systematically dismantled.
The BJP's final tally was 207 of 293 seats with 45.84% of the vote; TMC won 80 seats with 40.8%. A five-percentage-point difference in vote share determined whether West Bengal — a state of 100 million people — would be governed by the BJP for the first time in its history. In a state where 9.1 million people were denied the opportunity to participate, the question of whether that five-point gap authentically reflects West Bengali preferences is not rhetorical. Al Jazeera's opinion section called it "a dangerous rupture in the democratic compact." I'd go further: when the number of excluded voters exceeds the winner's margin in one-sixth of all constituencies, describing the result as a legitimate democratic mandate requires a definitional flexibility I'm not prepared to extend.
Positive & Negative Analysis
Positive Aspects
- Global Alarm About Algorithmic Voter Suppression Has Reached Critical Mass
One of the few genuine positives emerging from this deeply troubling episode is the level of international attention it has generated around a threat that most democratic governments were still treating as theoretical. CNN, Al Jazeera, Democracy Now!, and the Christian Science Monitor all ran major investigations, giving the story a global audience that state-level Indian elections rarely receive. The Brennan Center for Justice published a landmark report specifically on preparing for AI-backed voter suppression, citing India alongside Wisconsin's 2024 EagleAI-assisted deletion of 166,433 voters as proof that the threat is already operational in multiple jurisdictions. What this media and research attention accomplished is a category shift: "algorithmic disenfranchisement" moved from a speculative future risk to a documented present reality in the minds of policymakers, civil society organizations, and election administrators worldwide. Policy responses to concrete, documented harms move faster than responses to hypothetical risks, and India's experience has provided the concrete documentation that the field needed. I'd argue that West Bengal's democratic tragedy has functionally served as the alarm that may prevent similar systems from being quietly deployed in other electoral contexts without scrutiny — a deeply uncomfortable kind of value, but a real one.
The international conversation now includes a specific, actionable demand: election algorithms must be classified as high-risk AI systems subject to mandatory pre-deployment auditing, transparency disclosure requirements, and meaningful human override mechanisms. The EU AI Act's ongoing revision discussions have been directly influenced by this case, with advocates pushing to ensure that electoral AI joins facial recognition and credit-scoring algorithms as a category requiring documented impact assessments. That regulatory momentum — however incomplete and slow-moving — represents a measurable global good that emerged directly from West Bengal's crisis.
- Civil Society Resistance Demonstrates That Democratic Institutions Are Not Yet Dead
Despite the scale of what happened, India's civil society response has been both rapid and substantive, suggesting that the democratic immune system has not been fully neutralized. The People's Union for Civil Liberties (PUCL), the Association for Democratic Reforms (ADR), Article 14, The Wire, and a coalition of international human rights organizations mobilized within weeks of the SIR becoming publicly known. ADR's Supreme Court petition, which completed oral arguments in January 2026, represents a serious legal challenge that keeps judicial remedy theoretically available. The Supreme Court invoked Article 142 of the Constitution during the election period to order the issuance of some supplementary voter list updates — a rare emergency intervention that established a precedent for judicial oversight of SIR going forward. The Indian American Muslim Council's request for an investigation by UN Special Rapporteurs has opened an international accountability channel, even if it currently lacks enforcement teeth. TMC MP Mahua Moitra's international networking, combined with The Wire's partnerships with global media, has ensured that this story remains alive and traceable — not something that can be quietly dismissed as routine Indian political turbulence. The breadth of this coalition, spanning domestic legal advocacy, international human rights mechanisms, and investigative journalism, represents genuine organizational capacity that matters for what comes next.
- Independent Investigative Journalism Has Proven It Can Still Find the Truth
The Wire, Scroll.in, AltNews, and Article 14 deserve specific credit for the quality of their investigative work under conditions that included RTI requests being stonewalled or answered incompletely. AltNews built the dataset proving that deleted voter counts exceeded victory margins in 49 constituencies — not a simple analysis when the government refuses to release the underlying SIR data. Scroll.in expanded that analysis to demonstrate the same pattern held in half of all BJP-winning seats. These achievements required cross-referencing Election Commission results with partial government disclosures, court documents, SABAR Institute reports, and constituency-level field interviews. That this kind of forensic data journalism remains possible in India is genuinely meaningful, because authoritarian consolidation relies fundamentally on information opacity. The fact that independent Indian media and international outlets like AFP and Al Jazeera were able to quantify what happened in West Bengal — with enough specificity to withstand expert scrutiny — before the news cycle moved on demonstrates that India's democratic information ecosystem has not been fully captured. This journalistic capacity is not incidental to democratic recovery; it is one of its essential preconditions.
- Voter Turnout Data Suggests Democratic Resistance Is Already Building
One of the more counterintuitive signals in this story involves what happened to voter behavior in constituencies with heavy Muslim populations. In 32 Muslim-majority constituencies, TMC's vote share dropped by 16 percentage points — but overall turnout in those same constituencies rose by 7.6%. This is a remarkable and important data point. Voters who were not deleted from the rolls showed up in greater numbers even as their community's traditional party was losing ground, suggesting that the SIR process itself galvanized political engagement rather than suppressing it. The most plausible interpretation is that voters who remained on the rolls understood what was at stake and voted accordingly, while the political realignment away from TMC indicates an active search for new political vehicles rather than passive resignation. Historically, voter suppression campaigns tend to generate short-term electoral benefits for the suppressing party but medium-to-long-term backlash as targeted communities build organizational capacity, cross-party coalitions, and voter registration infrastructure of their own. The 7.6% turnout increase is, for me, the most hopeful number in this entire analysis — it suggests that democratic will survives even when democratic process is being systematically corroded.
- The Case Is Accelerating Concrete Progress on Global AI Governance
West Bengal's voter purge has functioned as a policy accelerant at the global level in ways that were not happening before the scale of the SIR became publicly documented. The February 2026 New Delhi Global AI Summit, attended by representatives from over 100 countries, included pointed criticism that the "Inclusive AI Development Declaration" signed by 88 nations contained no specific provisions on electoral AI bias — a critique that has since generated a formal working group on the topic. The Brennan Center's framework for AI-backed voter suppression defense is being actively referenced by election administrators in the United States, Germany, and Brazil as they evaluate their own voter list management systems for analogous risks. India's case is being cited in academic and policy literature as the foundational empirical evidence that electoral AI requires dedicated regulatory treatment rather than coverage under generic AI governance frameworks. I believe this kind of regulatory cross-pollination — where one nation's documented failure creates the evidence base that drives other nations toward precautionary regulation — is one of the more important global goods to emerge from an otherwise catastrophic situation.
Concerns
- SIR Phase 3 Is Already Scheduled to Multiply the Damage Fourfold
The most urgent concern right now — as of May 2026 — is that the process that deleted 9.1 million West Bengal voters is not being corrected. It is being expanded. SIR Phase 3, officially announced on May 14, 2026, covers 16 states and three union territories including Delhi and Maharashtra, targeting 367.3 million voters between May 30 and October 14. If the same 10% deletion rate applied in Phases 1 and 2 holds, approximately 36.7 million additional voters could be removed from the rolls — four times the West Bengal figure, in states that include India's largest urban populations and most competitive electoral battlegrounds. The timing is not coincidental: Phase 3 falls directly before the 2027 state assembly elections in several of these states. The corrective lessons from West Bengal — about transliteration errors, about women's name changes at marriage, about the structural impossibility of processing millions of objections in weeks — are not being incorporated into Phase 3's design. They are being overridden by the political logic of scaling a system that delivered results in West Bengal. If Phase 3 proceeds without judicial intervention, what happened in West Bengal will look, in retrospect, like the pilot study. I find that genuinely alarming.
- Full Algorithmic Opacity Makes Accountability Structurally Impossible
Citizens and activists who filed Right to Information requests seeking documentation of the SIR process — the necessity assessments, the algorithmic criteria, the internal approval chain — received what investigators described as "incomplete answers or outright refusals." India's Digital Personal Data Protection Act 2023 contains no algorithmic accountability provisions whatsoever and, unlike the EU AI Act, imposes no obligation on any government body to explain automated decisions to individuals affected by them. NITI Aayog's National AI Strategy explicitly identified caste bias and language bias as AI risk categories requiring attention — but electoral algorithms were notably excluded from that classification, a gap that reads as intentional given the political stakes involved. The February 2026 New Delhi AI Summit concluded without binding commitments on electoral AI bias despite a vocal chorus of civil society organizations demanding exactly that. The fundamental contradiction here — a country signing "inclusive AI development" declarations while simultaneously running an undisclosed algorithm that has removed tens of millions of citizens from the right to vote — is not a paradox that can be resolved through better communication. It reflects a deliberate choice to maintain opacity precisely because transparency would make the system's operation politically untenable.
- International Accountability Mechanisms Have Proven Structurally Toothless
The absence of effective external accountability for India's democratic backsliding is one of the most consequential structural failures in this entire story, and it is not an accident. The European Union imposed legal sanctions and budget penalties on Hungary and Poland for democratic regression far less severe than what India's SIR represents. For India, the EU has maintained what ECPR's The Loop accurately called "strategic silence" — driven by the EU-India Free Trade Agreement being described internally as "the mother of all deals," a negotiation so economically significant that human rights conditionality has become politically untenable. The United States faces its own contradiction: India is a cornerstone of the anti-China strategic partnership in the Indo-Pacific, making sustained democratic criticism toward New Delhi fundamentally incompatible with Washington's regional security architecture. The IAMC's UN Special Rapporteur request carries no binding force. What this means in practice is that the combination of economic interest (EU) and security interest (US) has effectively immunized India from the kind of external pressure that has historically served as a meaningful corrective when domestic institutions fail. External pressure combined with internal institutional compromise — a judiciary under pressure, an Election Commission captured — means that the normal feedback loops that slow democratic erosion are not functioning.
- Women Voters Face a Compounding Double Discrimination
That 62% of deleted voters in West Bengal were women is a fact that has received far less attention than it deserves, and the mechanism is specific and damning. Muslim women in West Bengal often have their names registered differently before and after marriage — a cultural practice that the SIR algorithm classified as evidence of two distinct individuals, triggering deletion for the post-marriage identity. This is what scholars of algorithmic harm call intersectional discrimination: the system is biased against Muslims and biased against women whose naming conventions deviate from majority-group patterns, and these two biases compound on the category of Muslim women, who face both simultaneously and at amplified intensity. Civil rights attorney Darshana Mitra stated the implication directly: "The government continues to use these errors — it shows that it is either mindlessly done out of lack of care, or it is designed to exclude." The Supreme Court's own acknowledgment that the SIR criteria don't reflect Indian social realities — including early marriage and late childbirth — transforms the government's continued use of these criteria from a matter of ignorance into a matter of choice. An already underrepresented demographic in Indian electoral politics is being further systematically removed from democratic participation, and the story is being underreported precisely because it sits at an intersection of gender and religion that mainstream political coverage tends to flatten.
- Democratic Indices Are Falling and Dragging the Economy Down With Them
Freedom House has now downgraded India by a cumulative 14 points since 2005, classifying it as "Partly Free" with a score of 62 out of 100 after stripping its "Free" designation in 2020. V-Dem ranks India 105th of 179 nations in its Liberal Democracy Index — down from 100th the previous year — maintaining the "electoral autocracy" classification it first applied in 2017. BTI 2026 scores India's democratic transition at 6.10 out of 10, placing it at the edge of "limited democracy." These are not symbolic designations — they correlate with measurable economic outcomes that are already showing up in the data. India's net Foreign Direct Investment in FY2025 collapsed to $353 million, down 96.5% from $10 billion the previous year, as foreign companies repatriated $49 billion in capital. Multiple analysts, including BTI, connect governance uncertainty directly to this investment retreat. India was widely understood as the leading alternative to China as the world's next major manufacturing hub — a destination for supply chain diversification driven by geopolitical risk-reduction logic. Sustained democratic erosion, recurring electoral legitimacy crises, and the institutional uncertainty generated by SIR are making that calculus more complicated with each passing quarter. The economic cost of authoritarianism is never immediate — it accumulates gradually, in cancelled investments, in talent emigration, in the compounding discount rate applied to political-risk assessments. But the FY2025 FDI data suggests the bill has already begun to arrive.
Outlook
In the short term — the next one to six months — two variables will define India's democratic trajectory, and they are racing against each other in real time. The first is the Supreme Court's pending verdict in ADR v. ECI. Oral arguments concluded on January 29, 2026, and no ruling has been issued. If the Court finds SIR unconstitutional under Article 326's guarantee of universal adult suffrage and orders Phase 3 halted, it would represent the most significant democratic correction India has seen in years. I want to be direct about how likely I think that is: not very. BTI 2026 documents that the Supreme Court has "repeatedly yielded to pressure from the Hindu nationalist government," and the court faces a backlog of 44 million pending cases — including this one, which has already been sitting for months without a decision. The Court did invoke Article 142 during the West Bengal election to order supplementary voter lists in specific constituencies, which proves that the institutional capacity for intervention exists. But capacity and willingness are different things, and the political weight pressing against a ruling that would effectively invalidate West Bengal's results and halt Phase 3 is substantial.
The second variable is Phase 3 itself, which has already been officially announced with a May 30 start date covering 367.3 million voters across 16 states. This is the more consequential variable because it is already in motion. If the Supreme Court verdict arrives after Phase 3 begins — which is the more probable sequence — judicial relief becomes retrospective at best and structurally meaningless for the tens of millions who will have already been deleted. The lesson of West Bengal has not been incorporated into Phase 3's design. The same algorithm, the same criteria, the same objection-processing infrastructure that handled 0.05% of West Bengal's appeals — all of it is scheduled to run at four times the scale in India's most populous urban states. The window for preventing what West Bengal experienced from becoming the permanent national norm is, right now, extremely narrow. I think that's the most urgent thing to say about the short term.
In the medium term — six months to two years — the most consequential variable is the fate of the "One Nation, One Election" legislation. The BJP Cabinet approved the bill in September 2024, and it was submitted to the lower house in December 2024. ONOE would synchronize all Indian elections — national and state — into a single simultaneous cycle. Think through what that combination means when placed alongside SIR: a government that controls the ECI could run a single nationwide voter list purge immediately before a synchronized election, rather than having to manage the political and logistical complexity of separate state-level purges on different timelines. The structural advantage this would give the ruling party is difficult to overstate. Currently, the BJP holds 240 of 543 lower house seats, short of the two-thirds majority required for constitutional amendment. But coalition arithmetic and the political momentum of West Bengal's historic BJP win make that threshold less stable than it might appear. If ONOE passes alongside continued SIR expansion, India's federal democratic architecture — historically the main structural counterweight to central power through the rotation of state governments — would face its most serious challenge since Indira Gandhi's 1975 Emergency.
The economic dimension of democratic backsliding deserves more analytical attention than it typically receives in political coverage of India. The FDI collapse — from $10 billion net inflow in FY2024 to $353 million in FY2025, driven by $49 billion in foreign capital departures — is not happening in isolation from the governance story. It reflects a recalibration of long-term investment risk assessments by multinational firms that must factor institutional stability, rule of law, and electoral predictability alongside market size and labor cost. India was broadly understood as the primary alternative to China for supply chain diversification — a destination that offered democratic governance as a distinguishing feature from its geopolitical rival. Sustained electoral legitimacy crises and SIR-generated governance uncertainty are corroding that differentiator. The economic cost of authoritarianism is never immediate — it compounds gradually, in the discount rates applied to institutional risk, in cancelled manufacturing investment decisions, in the brain-drain arithmetic of talented workers choosing Singapore or Dubai over Delhi or Bangalore. The FY2025 FDI data suggests that compounding has begun.
Looking two to five years out, three scenarios bracket the plausible range of outcomes. The bull scenario — which I assign roughly 15% probability — requires the Supreme Court to deliver a decisive ruling against SIR's constitutional basis, triggering legislative reform of the 2023 ECI Appointment Act and, eventually, meaningful restoration of electoral commission independence. In this scenario, the 7.6% turnout increase among Muslim voters in West Bengal's Muslim-majority constituencies — the clearest evidence that suppression can galvanize rather than demobilize political participation — becomes the foundation of a broader opposition coalition capable of winning seats in 2027 and 2029. International pressure, particularly if EU-India FTA negotiations are conditioned on rule-of-law benchmarks under European Parliament pressure, adds external momentum to internal recovery. Global AI governance frameworks converge on mandatory pre-deployment audits for electoral algorithms, creating normative pressure on India even absent direct sanctions. I want to be honest: I think this scenario is possible but not probable. The institutional damage already done — a captured Election Commission, a judiciary under documented pressure, a civil society fighting uphill against an information environment shaped by state-aligned media — is not reversed by a single court ruling. Democratic recovery from "electoral autocracy" designation, per V-Dem's historical data, is among the rarest political transitions on record.
The base scenario — roughly 55% probability — is continuation of the current trajectory. SIR Phase 3 runs as scheduled. The Supreme Court delivers a verdict too narrow or too delayed to halt Phase 3's effects. The 2027 state elections and the 2029 general election proceed under the same algorithmic framework, with the same ECI institutional structure, in a country where electoral legitimacy is contested but elections continue to be held and results accepted by international observers under a diplomatic silence driven by strategic interests. India's Freedom House score continues declining toward the 55-60 range, and V-Dem's ranking drops into the 110th percentile. International partners maintain their strategic silence, calculating that India's geopolitical value as a counterweight to China outweighs the reputational cost of normalizing its democratic regression. Elections occur, parliament convenes, opposition politicians file legal challenges — but the structural preconditions of meaningful electoral competition are progressively hollowed out. This is what political scientists call "competitive authoritarianism" or "electoral autocracy": the forms of democracy maintained while its substance is systematically drained. It is, in my judgment, where India currently sits most firmly on the trajectory map.
The bear scenario — roughly 30% probability — requires ONOE passage combined with SIR nationalization, completing what would effectively be a one-time institutional lock-in of BJP structural electoral advantage. In this scenario, the Election Commission becomes functionally indistinguishable from a ruling-party auxiliary, and every election cycle begins with a nationwide voter list purge calibrated to the political geography. India's Freedom House score drops below 50 — the "Not Free" threshold — a reclassification that would mean 1.4 billion people moving from a "Partly Free" country to a "Not Free" one. Freedom House data shows "Not Free" countries have grown from 45 to 59 between 2005 and 2026 — a 31% increase in 20 years. India's addition to that category would push approximately 90% of the world's population into either authoritarian or "not free" regimes. V-Dem's 2026 report notes that freedom of expression is "the most common target among autocratizing leaders over the past 25 years" — intensified pressure on The Wire, Scroll.in, AltNews, and Article 14 in this scenario would accelerate the timeline for full information-space capture. I assign this 30% probability not because I think it's unlikely in the abstract, but because the civil society evidence — the AltNews investigations, the ADR litigation, the turnout surge — suggests India has not yet exhausted its democratic immune response.
Let me be candid about where my analysis could be wrong. India's civil society has surprised observers before, and the independent media infrastructure has survived pressures that would have extinguished comparable outlets elsewhere. Economic pain from FDI collapse may force a BJP government that prizes growth above most other metrics to recalibrate its governance posture faster than its political instincts would otherwise dictate. International normative pressure, though currently toothless, could harden into trade conditionality if EU-India FTA negotiations reach a stage where the European Parliament inserts human rights benchmarks. And the sheer demographic weight of India's Muslim community — 200 million citizens — creates a political mobilization potential that has not yet been fully tested under sustained suppression conditions.
What I am confident about, however, is this: the question of whether India deserves to be called the world's largest democracy can no longer be answered by reference to its voter rolls or turnout figures. When 9 million citizens are deleted from the precondition of democratic participation by an algorithm the government refuses to disclose, when 99.95% of their objections go unprocessed, when the institution tasked with preventing this is appointed by the very party that benefits from it, and when international partners calculate that trade agreements and security partnerships outweigh the cost of speaking clearly — the word "democracy" is doing a great deal of work to cover a reality it no longer describes. An algorithm now decides who gets to vote in the country of 1.4 billion people most frequently cited as proof that democracy scales. That should alarm every democracy on earth, because the next version of this algorithm is probably already being evaluated somewhere else.
Sources / References
- Muslims the target? Fury as millions lose voting rights in India's Bengal — Al Jazeera
- The BJP's Bengal victory exposes the erosion of Indian democracy — Al Jazeera Opinion
- India's Modi Gov't Purged Millions of Voters Before Elections in Direct Attack on Democracy — Democracy Now!
- India: Freedom in the World 2026 Country Report — Freedom House
- 2026 West Bengal Legislative Assembly election — Wikipedia
- Bengal results: In 49 seats, voters deleted after adjudication exceeded the victory margin — AltNews
- India voter register deletions spark outrage over AI use and poll fairness — The Star/AFP
- Democratic Backsliding in India: Electoral Manipulation and the Deterioration of Civil Liberties — Democratic Erosion (Brown University)