A formal complaint has been lodged with the Chief Electoral Officer of Telangana, alleging that photographs collected for voter ID cards were illegally shared with the state government and repurposed to build a sweeping facial recognition system.
Independent researcher Srinivas Kodali, who filed the complaint, told BOOM that the images collected solely for elections are now being fed into RTDAI (Real Time Digital Authentication of Identity)—a facial recognition system already in use across government departments and tested during municipal polls. The allegations raise serious questions about legality, privacy, and voter rights.
Kodali has been tracking the misuse of electoral rolls since 2015, when Aadhaar and voter IDs were briefly linked for de-duplication before the Supreme Court ordered the practice to stop. By 2020, when Telangana formally announced its facial recognition system, he had already filed a case in the Telangana High Court against the Election Commission of India for sharing voter data with the state.
“SQ Masood, another activist, later filed an RTI with the government. The documents he obtained confirmed that electoral photographs are being used for facial recognition in welfare schemes and governance,” Kodali told BOOM.
What is RTDAI?
At the centre of the controversy is RTDAI—Real Time Digital Authentication of Identity—a facial recognition system developed by Telangana’s IT department in 2019 to make services “presence-less” and “contact-less”. It works by taking a live photo through a smartphone, verifying it for authenticity, and matching it against stored images, including voter ID photographs, along with basic demographic details.
Initially a pilot project for pensioners, who could prove their existence each year by simply clicking a selfie, the system quickly expanded. The Transport Department began using it for driving licence renewals, the Education Department applied it during college admissions, and most controversially, it was tested during municipal elections to verify voters at polling stations.
While the government presents RTDAI as a tool of convenience and efficiency, the complaint argues that its foundation rests on electoral photographs that were never meant to be used outside elections, raising significant concerns about consent, legality, and privacy.
How Voter Data Ended Up in RTDAI
As mentioned in the complaint, the RTI documents obtained by Kodali and SQ Masood over a couple of years reveal the path of voter data into RTDAI.
The RTI document from Telangana government confirm that voter photographs collected for Electoral Photo Identity Cards (EPIC) were accessed and used in the state’s facial recognition system.
Similarly, the RTI from the Election Commission also admit that during the 2015 National Electoral Roll Purification–Authentication Programme (NERP-AP)—a process designed to update and clean voter lists in Andhra Pradesh and Telangana—all EPIC data was shared with Telangana’s State Resident Data Hub.
Although the Supreme Court had ordered NERP-AP to stop, by the time the instructions reached the Election Commission, the state government had already received and used the data. Several voters were removed from rolls without due procedure, affecting the 2018 Telangana Assembly elections.
Legal and Privacy Concerns
Kodali’s complaint highlights the serious implications of repurposing voter data. Under the Representation of People’s Act, 1950, electoral rolls are meant solely for conducting elections, and sharing voter ID photos with state agencies for unrelated purposes is not authorised by law.
Voters provided their photographs for elections, and using them for RTDAI without consent raises fundamental privacy concerns. The Supreme Court’s landmark Puttaswamy judgment in 2017 recognised privacy as a constitutional right, yet this principle appears to have been overlooked.
Annexure documents from the Telangana government illustrate how the facial recognition system functions, highlighting both its capabilities and limits. In “high-confidence” matches, the system can identify the same person with similarity scores above 90%. It can also correctly reject mismatches, showing low similarity scores for different people. Yet, failures are evident: in some cases, the same person is not recognised, with scores as low as 34–50%.
These examples underscore the fallibility of facial recognition technology. Misidentifications or failed matches could deny genuine voters their right to vote, while the reuse of electoral data for pensions, education, and transport points to a creeping surveillance infrastructure. In the absence of comprehensive data protection laws, there are few safeguards to prevent governments from repurposing citizens’ data, eroding public trust in elections and governance.
Expansion Beyond Telangana
Kodali cautions that such invasive systems are already spreading beyond Telangana. “In Telangana, this system was piloted in mock smartphone-based voting,” he said, adding that similar implementations have now appeared in Bihar’s municipal elections, even without laws governing its use.
He said that if the Election Commission does not act, he plans to file fresh litigation targeting the use of facial recognition in elections. He explained that while governments are technically exempt from data protection laws, this does not give them free rein to bypass the Supreme Court’s ruling on the fundamental right to privacy.
“The more people understand what is happening, the more they will demand accountability from the government,” Kodali said, emphasising the importance of public awareness in curbing misuse of citizen data.