Website maps voter data, family, neighbors

A website is offering access to digitized Election Commission of India voter rolls containing details for approximately 57 crore electors. The service, which MediaNama was pitched via email, allows queries on individuals, households, and booth-to-constituency composition through a custom site, data extract, or API.

From a single name search, the website revealed there are 20 voters in India with the same name as the MediaNama reporter, 12 in his state, and 2 in his district. The AI-built platform also generated an extensive relationship graph. It identified relatives and generational links, listed 26 other people at the same neighborhood address, and disclosed who lives in adjacent houses or blocks.

Privacy risks of electoral data mapping

A detailed relationship graph for any individual can be created, encompassing immediate family, neighbors, and their associated voting booth. Mapping political data is not new. In 2013, Netcore founder Rajesh Jain, then a key member of Narendra Modi's team, explained the process at an IAMAI summit.

Jain described how combining a voter ID number from SMS responses with electoral roll data provides a "unique geo-identity." This double KYC (Know Your Customer) method, using the voter ID and mobile number, allows for confirmation and continuing engagement, enabling micro-targeted political campaigns.

Digitization transforms a manual, constituency-specific lookup into a nationwide, fuzzy search capable of instant individual reports. Several critical privacy issues arise from this capability.

Mass search and surveillance becomes possible, especially when layered with other datasets. Household and social-graph reconstruction uses AI to create embeddings that expose relatives, surnames, house numbers, and adjacent households, linking them together without consent.

Political and community profiling can occur by layering historical voting data over this information. Polling station geography, surname, and household metrics can become unreliable proxies for caste, religion, income, or political preferences, enabling microtargeting, discrimination, and communal targeting.

Enabling fraud and data scraping

The system is also enabling fraud, particularly social engineering. The Indian government's focus on measures like SIM binding does little to address the rampant disclosure of personal information that fuels such fraud. This specific data mapping, done for political purposes, could worsen fraud through relationship mapping.

Furthermore, the platform enables reuse and AI scraping. It generates a JSON file for each individual, and the service pitch included offering API access. This removes the need for others to perform OCR, deduplication, or relationship mapping themselves. Although the site uses a robots.txt file with a noindex clause to prevent search engine scraping, this is insufficient. MediaNama has experienced Chinese AI bots and Amazon's scraping bots ignoring such directives, leading to server issues.

How the AI system was built

An analysis of the JSON file output for the reporter's data suggests the system used one Claude Haiku 4.5 call, processing 152 input tokens and 27 output tokens at a cost of ₹0.13. The process to achieve a confidence score for individuals involved 50 searches that screened 62 records.

The system queried all nationwide name matches, comparing exact-phrase versus "sounds like" matches. It examined state and district subsets, name plus relative-name pairs, and candidates within a three-year age tolerance. It also analyzed reordered name tokens, shortened and full versions of relative names, addresses associated with possible relative matches, and geographical and roll-type facets.

Search modes visible in the JSON include all, phrase, and any. The engine begins with a name-frequency prior and multiplies prior probabilities by hard-coded "likelihood ratio" weights. For household and neighbor links, it queries everyone sharing a rollId, partNumber, and normalized house field.

For family relationships, it connects an elector to a named relative like a father, husband, or mother. The engine creates a node for the elector, searches for the named relative as another elector, and creates a "ghost" node if no confident match is found. It assigns a generation from the relation type and attempts to merge nodes using names, sex, age gaps, and co-residence. It also looks for shared named relatives across nearby households.

While a fundamental right to privacy exists, there appears to be no requirement for state bodies like the Election Commission of India to avoid doxxing citizens. The Government of India refuses to enact surveillance reform, continues to weaken the RTI Act to protect the bureaucracy, and exempts government bodies from laws that should apply to them. Courts rarely address these concerns.