Vijayawada, India – As soon as Meda Ramana was elected the head of Garapadu village council in the southern Indian state of Andhra Pradesh, several residents complained to her of being wrongfully excluded from government welfare schemes.

About 50 villagers – mostly women of the Dalit, Indigenous and religious-minority communities – told her in early 2021 that their annual welfare benefits, of 10,000 rupees ($119) to 120,000 rupees ($1,430), under various schemes, had been abruptly stopped by the state government.

At first glance, it seemed to be an innocuous data entry error. But when Ramana enquired about the reasons, staff at the village secretariat told her the villagers were ineligible for the welfare because they had “migrated”. Ramana told the secretariat that the complainants were very much living there, but to no avail.

She then complained to higher officials in the district administration who ordered the village staff to conduct a probe. But the village staff ignored those directives. It was at this point that Ramana began to smell a political ploy.

Andhra Pradesh is one of the two states that held regional elections in May 2024, during the time of India’s national poll that re-elected the Narendra Modi-led Bharatiya Janata Party (BJP). The Andhra Pradesh state election was fought between two dominant regional parties – the incumbent YSR Congress Party (YSRCP) and the Telugu Desam Party (TDP).

The TDP won, but in the run-up to the elections, the outgoing YSRCP government ran a sophisticated, data-driven voter influence campaign that wrongfully denied poor and vulnerable communities their welfare benefits because members of those communities were likely to vote for the opposition, our investigation reveals.

Multiple media reports have highlighted how governing parties in India – the BJP at the national level and others in the states – have accessed and misused government-collected personal data of citizens in past elections. The TDP did the same when it was in power in Andhra Pradesh during the previous election in 2019.

Our investigation reveals, for the first time, the real-world harm caused by such misuse – with the vulnerable being denied their rightful welfare benefits and their rights to exercise free choice in elections undermined.

Welfare delivery was a major election plank in Andhra Pradesh. The then-incumbent YSRCP government said it transferred about 4.3 trillion rupees ($50.7bn) in its five-year term to about 127.4 million poor and marginalised people, including through pensions and housing and food support.

It appointed paid “volunteers” to “proactively” deliver welfare at citizens’ doorsteps, replacing the traditional method in which elected village councils, known as panchayats, administered the welfare.

One volunteer was appointed for every 50 households – approximately 260,000 volunteers covered more than 1.3 million households in the state. Every volunteer was paid 5,000 rupees ($60) per month from the state exchequer.

The YSRCP government also gave this army of volunteers unfettered powers and technological tools to indiscriminately harvest the personal data of citizens. It used the data to create voter “profiles” and predict their political choices.

The information was then used by the government volunteers and YSRCP cadres to influence these voters including through dubious tactics like excluding opposition-inclined voters from welfare schemes.

When Ramana and her husband looked through the list of Garapadu’s excluded families in 2021, they saw a pattern. “All the wrongly-excluded families were supporters of TDP”, the opposition party at the time, claimed Sagar, Ramana’s husband who is also an active TDP worker.

The couple concluded that the “village volunteers”, were singling out TDP supporters and cutting off their benefits “to pressurise them to switch their support to YSRCP”.

In April 2022, 27 women from Garapadu filed a case in the state’s high court against the village volunteers and government officials for “illegally” cutting off their benefits for “political reasons”. The court found the women were “eligible” for the schemes and reprimanded the state for the village volunteers’ “illegal actions”.

The women’s benefits were resumed after the court’s order and residents of several other villages from the state filed more than 100 similar complaints in court the same year, said G Arun Showri, a lawyer who represented the Garapadu petitioners in the high court.

“All claimed the village volunteers blocked welfare benefits of eligible claimants who were supporters of the opposition party,” Showri said.

But how did the volunteers know the political preferences of villagers? A review of the volunteer scheme’s documents – service contracts, source code of the software application they used, official WhatsApp groups – and more than a dozen interviews with political workers, campaign managers and voters, revealed the method.

Predicting voting choices

The volunteers were required to conduct a “baseline survey” of the assigned households using a mobile application. An analysis of the source code of the app, first carried out by internet-governance researcher Srinivas Kodali, shows data to be collected for every resident – including their home address, employment and family information such as data on caste, religion, education and health status (with granular details such as diseases and pregnancies).

These were meticulously recorded and constantly monitored. The source code analysis was reviewed and verified by this correspondent.

The volunteers conducted periodic surveys, including over WhatsApp, to keep the profiles updated. Each volunteer was allocated a WhatsApp group of 50 households. The volunteers recorded data on the “felt needs” of the citizens and their “feedback” about the governance.

“Volunteers will identify the problems being faced by citizens in their jurisdiction, and the same will be brought to the notice of government officials to get them resolved,” said the government order for recruiting volunteers.

Sagar, Ramana’s husband, said: “Because of this persistent monitoring of households over five years in the name of government surveys, the volunteers have become aware of the complete profile of the households, including their likes and dislikes, mental stresses and financial problems. They can then easily assess which household is a supporter of the YSRCP government, the opponents and neutral voters.”

Bhaskar Basava, a journalist in Andhra Pradesh, said the YSRCP had formed committees at every polling booth to segregate the voters into “YSRCP-inclined”, “opposition-inclined” and “neutral” categories, with the assistance of the volunteers. “They even colour-coded the voters in the voter lists to mark these categories,” he recalled.

N Ramesh Kumar, a former Andhra Pradesh election commissioner and secretary of the nonprofit Citizen for Democracy, said the scheme was “one of the most sophisticated, ubiquitous and far-reaching data-driven voter manipulation campaigns ever designed by a political party – far bigger than Cambridge Analytica.” He was referring to the United Kingdom-based political consultancy that mined the social-media data of voters in the UK and the United States to predict their political inclinations.

The Andhra Pradesh government volunteers “combined intensive door-to-door surveys with government databases to create exceptionally accurate voter profiles”, Kumar explained.

Personal data Wild West

Political parties across the world collect data on voters’ needs and preferences through surveys to target them in campaigns. But when citizens’ data, collected through the machinery of government, is accessed and misused by the governing party, it distorts the level playing field by giving the party an unfair advantage.

India has not enacted a personal data protection law that protects the misuse of citizens’ data by state and non-state actors, despite the country’s Supreme Court in 2018 directing the government to do so.

Kodali, the internet governance researcher, calls the village volunteers’ operations in Andhra Pradesh “a classic case of cybernetics”, referring to the science of communication and control in machines and humans.

“Any organised actor can collect loads of personal data of citizens in India because there is no law to prevent it …  Control over people is exercised when data leads to exclusion or fear of exclusion among people. Through welfare exclusions of the poor, YSRCP precisely tried to achieve that,” Kodali told Al Jazeera.

Tejasi Panjiar, a former associate policy counsel at the New Delhi-based nonprofit Internet Freedom Foundation, said such operations “would be held illegal in other democracies such as the EU and Australia that have strict data protection laws”.

“This kind of function creep, when citizens allow the government to collect their data for one purpose [like welfare delivery] but the data is misused for another purpose, is against the principles of data protection and violates citizens’ right to privacy.”

YSRCP president and former Chief Minister YS Jagan Mohan Reddy’s office did not respond to email queries. The party’s national spokesperson, K Ravi Chandra Reddy, said he was not acting on behalf of the party “at the moment”.

Designed to misuse

The volunteer scheme was executed by the YSRCP government in consultation with the election management consultants hired by the party.

The YSRCP government recruited volunteers from each village based on interviews by an ad hoc committee of local government staff, and not through a centralised recruitment exam, the scheme documents show.

“Only the names recommended by local YSRCP leaders would get selected,” Ramana said. Several party workers Al Jazeera spoke with echoed this.

The government then created a parallel system of governance, managed and run by election strategists who closely worked with it, to oversee and direct the volunteers’ work.

The government floated a tender in November 2019 to hire a field operations agency (FOA) to “design, develop and manage” a robust technology-enabled application to “gather feedback on service delivery and collect people’s aspirations and grievances on a continuous basis” and analyse the data collected by the volunteers.

The contract was bagged by a consortium of companies led by Raminfo Ltd. The contract was initially for one year until March 2021 at a cost of 686 million rupees ($8.1m) and was later extended for another year at an additional 675 million rupees ($7.9m). The consortium was also tasked with developing and maintaining a data portal for volunteers until March 2024.

The FOA, which on its X account described itself as a “liaison body between the government of AP [Andhra Pradesh] and the village volunteers”, promoted YSRCP from its now-deleted X (formerly Twitter) account, including posts on how the citizens of AP wanted YSRCP to come back to power. Al Jazeera accessed the FAO’s deleted tweets using Wayback Machine.

To monitor the volunteers’ work in the field and coordinate with the government departments, the FOA hired large numbers of staff, according to the tender document, including at the state, district and mandal levels, creating a chain of command in that order. (Mandal is an administrative unit between village and district levels.)

LinkedIn profiles of several FOA coordinators show they were hired from a political consultancy firm called the Indian Political Action Committee, or IPAC, an old strategist for the YSRCP.

“The FOA and IPAC worked closely with the YSRCP leaders to promote the party. It was the seamless integration of all from the state level to village level. There was no difference between the government programmes and the party programme,” said Kumar of Citizen for Democracy.

Raminfo said “as FOA was an e-Governance project, we participated in the tendering process and completed the project much prior to the elections.”

“Our scope is limited to the training of volunteers on schemes/programs developed by departments. Volunteers are appointed and managed by the Government. We didn’t build any apps or software for the volunteers’ duties or operations. Our contract/agreement/relationship is only with the user department which floated the RFP,” the company added.

An IPAC spokesperson said in a WhatsApp text that “IPAC doesn’t have access to/nor do they use voter data”.

Seamless integration

In response to a Right to Information request, the state government claimed that all the data collected by the volunteers’ app was secured in the state’s data centre and was only shared with the government departments after signing “consent/access forms” with the respective department, without giving more details.

In practice, however, nothing stopped volunteers from sharing their granular insights on each household they tracked with the local YSRCP leaders.

A former volunteer in Garapadu, who spoke on condition of anonymity, said the data collected from all the surveys conducted by the volunteers was first shared with the local YSRCP leaders before it was submitted to the officials responsible for welfare schemes.

“Local YSRCP leaders would tell us to manipulate the beneficiaries lists and remove TDP supporters and neutral villagers from them. We would then tell those villagers to join the YSRCP party to avail benefits of the schemes,” he said.

For those beneficiaries who received welfare benefits, local YSRCP leaders had access to granular data of every penny they got, which they used in campaigns. Several posts during the election campaign and available on X show the party leaders handing pamphlets to individual families, addressed in the name of the head of the family, with details of all the benefits they received over the years, asking them to vote for the YSRCP.

The government-appointed volunteers, YSRCP’s local leaders and its election managers worked as well-integrated teams, as evident from this YouTube video released by a YSRCP-affiliated channel instructing the volunteers how they needed to campaign for the party. It asked volunteers to visit door-to-door with YSRCP leaders, survey the voter sentiments about the government and tell the voters how the YSRCP government was better than the previous one.

The integration of volunteers and party workers was also evident from the official WhatsApp groups created for the welfare schemes. All 13 volunteers in Garapadu village were managing WhatsApp groups that had at least one member from each household that was being tracked.

These 13 volunteers were part of another group that had district-level government officials and YSRCP leaders as members. That is where they “received information/instructions which needed to be circulated among villagers”, the former volunteer in Garapadu told Al Jazeera.

While officially, the WhatsApp groups were meant only for communication related to the government welfare scheme, the screengrabs of the groups show volunteers had sent messages promoting the YSRCP. Some of the messages in the groups – analysed by WhatsApp Watch, a platform by Princeton University’s Digital Witness Lab, which studies the use of WhatsApp in political campaigns – were based on misinformation that denigrated the TDP.

After it received several complaints of government-appointed volunteers campaigning for YSRCP, the Election Commission of India on March 20 filed criminal cases against them for “unduly influencing elections”. Ten days later, the commission barred all volunteers from carrying out their activities, including disbursing welfare benefits.

While the YSRCP eventually lost the elections, its misuse of personal data created a dangerous template for the state to discriminate against citizens based on their political views.

“Such granular profiling of voters by the state can lead to the end of the secret ballot. It will become very easy to predict who is going to vote for whom. And the [Andhra Pradesh] example shows how this information can be weaponised against the vulnerable,” said Shivam Shankar Singh, a former political consultant and author of the books, The Art of Conjuring Alternate Realities and How to Win an Indian Election.

With inputs from Prudhviraj Rupavath in Vijayawada.

(Kumar Sambhav is an independent journalist and India Research Lead with Princeton University’s Digital Witness Lab)