The State has found a new object to formalise: the Indian farmer. Through the Digital Agriculture Mission and its AgriStack architecture (a Farmer Registry, a Crop Sown Registry and geo-referenced village maps), the Union and the States propose to give every cultivator a unique digital identity. That identity will unlock institutional credit, crop insurance and procurement benefits. The economic logic is respectable. Formalising agricultural credit, displacing the moneylender and rationalising subsidy delivery are legitimate public ends.

The legal logic is another matter. The same architecture assembles, in one interoperable stack, a farmer's land holdings, cropping choices, plot coordinates and borrowing history. It does so for a population that is largely not digitally literate, often not documented as the actual cultivator, and in no position to negotiate the terms on which it is enrolled. The State's ambition is to see the farmer entirely. The farmer's legal ability to see what the State is doing with that sight is, at best, unaddressed.

The question is not whether AgriStack is useful. It is whether a system of this scale can lawfully operate without the consent mechanics, purpose limitation and accountability that the Digital Personal Data Protection Act, 2023 and the constitutional standard in K.S. Puttaswamy v. Union of India demand. On the present evidence, it cannot.

The Statutory Friction Point: Fiduciary or Public Authority?

A false dichotomy dominates this debate: that state agriculture departments are either "public authorities" discharging sovereign functions or "Data Fiduciaries" bound by statute. The DPDP Act does not recognise the first as a category of immunity.

Under Section 2(i), a Data Fiduciary is any person who, alone or in conjunction with others, determines the purpose and means of processing. "Person" in Section 2(s) expressly includes the State. A State agriculture department that decides which datasets populate the Farmer Registry, and for what purposes they will be used, is a Data Fiduciary. Because AgriStack involves the Centre, State agriculture and revenue departments and implementing agencies acting together, the "in conjunction" limb makes them joint fiduciaries. Section 8(1) makes each answerable for compliance irrespective of any inter-se allocation of roles. The departments cannot contract their way out of statutory liability.

The State's only route to escape is Section 17(2)(a), which allows the Central Government to exempt notified instrumentalities in the interests of sovereignty, security or public order. A farmer registry is not a security apparatus. Any such notification for AgriStack would fail the proportionality test laid down in Puttaswamy and applied in the Aadhaar judgment, which requires legitimate aim, rational nexus, necessity and a balance between the intrusion and the benefit.

The State's better argument is Section 7(b), which permits processing without consent for the provision of subsidies, benefits and services. That doorway is real, but it is narrow, and it does not admit what AgriStack's design contemplates.

Section 7(b) licenses data use to deliver the specified subsidy or benefit, not to build a persistent, multi-purpose farmer profile. The legitimate-use ground is anchored in Section 4 and carries the purpose-limitation logic with it. Nor can it carry onward disclosure to commercial actors. Sharing with an AgTech firm for credit scoring or marketing is not "provision of a subsidy." It is a new purpose, and it needs a new lawful basis, which can only be consent.

And consent, where it is required, must meet Section 6. It must be free, specific, informed, unconditional and unambiguous, given by clear affirmative action, with notice under Section 5 available in English or an Eighth Schedule language. A "consent" extracted by conditioning PM-KISAN or MSP access on enrolment is conditional by definition. Under Section 6(2), any part of a consent that infringes the Act is invalid to that extent.

The Act contains no "sensitive personal data" tier. Land title, plot coordinates, sowing patterns and credit history are each unremarkable in isolation. Aggregated and geo-tagged, they become a granular economic biography. The Act's only mechanism for scale-appropriate duties is Section 10 on Significant Data Fiduciaries, and the State has so far shown no appetite for subjecting itself to it.

When Error Becomes Entitlement

Indian land records are not titles. That has been the settled position since Suraj Bhan v. Financial Commissioner and Sawarni v. Inder Kaur: mutation entries and records of rights are fiscal documents, raising at most a rebuttable presumption, and confer no title. Yet the records are riddled with unmutated inheritances, stale entries, undocumented partitions and, above all, the invisibility of tenant and sharecropping cultivators, whose oral and informal tenancies appear in no register.

AgriStack takes this probabilistic, litigable, correctable document and uses it as the foundational layer of a deterministic digital identity. A rebuttable presumption becomes a machine-readable fact. The consequences follow mechanically.

The registered owner who is not the cultivator receives the credit, the insurance payout and the procurement eligibility. The actual cultivator, who is legally invisible, is excluded from all three. And downstream systems that treat the registry as authoritative (banks, insurers, procurement agencies) have no incentive to look behind it.

The Act formally provides a remedy. Section 8(3) obliges a fiduciary to ensure completeness, accuracy and consistency where data is used to make a decision affecting the Data Principal or is disclosed to another fiduciary, and Section 12 confers a right to correction. But a right to correct a data entry is not a right to adjudicate a title or tenancy dispute. The data layer cannot resolve what the revenue courts have not. The farmer must litigate in a forum with a different clock and different costs, while the digital record speaks with a confidence the underlying document never earned.

This is wrongful exclusion constructed into the system. It engages Article 14, because the classification between recorded owners and unrecorded cultivators is manifestly arbitrary, and Article 21, in its dimension of livelihood and dignity. The Aadhaar decision upheld the architecture while registering the Court's concern about exclusion. AgriStack presents the same concern at larger scale, with an even less reliable foundational dataset.

Corporate Exploitation Risk: Asymmetric Information, Inverted

Private participation was an early feature of AgriStack's design. The State's own pilots and memoranda contemplated technology and agribusiness firms working with the registries. Once those datasets are opened to private firms for algorithmic credit scoring and input marketing, the classical problem of asymmetric information, which Akerlof and Stiglitz and Weiss analysed from the lender's side, is inverted.

The traditional justification for rationing rural credit was that borrowers know more about their own risk than lenders do. A platform with land, crop, geo-location and repayment data eliminates that disadvantage for the lender. It does not, however, create symmetry. It creates a new asymmetry in the opposite direction: the farmer is now transparent, while the scoring model is opaque. The farmer cannot know what features drive a score, whether the inputs are accurate, or why credit was denied. The DPDP Act confers no right to explanation of automated decisions, and no right to contest profiling. Its Chapter on rights is thin precisely where an algorithmic regime needs it most.

The same data enables other forms of exploitation. Plot-level crop and acreage data lets pesticide and seed vendors time and tailor sales to a captive, informationally disadvantaged buyer. A seller who knows a farmer's acreage, crop cycle and credit stress can price accordingly. Aggregated sowing data, if it reaches traders, informs price expectations before harvest, to the detriment of the cultivators who generated it.

The liability gaps are real. A State-to-private disclosure requires its own lawful basis. If it rests on consent, the farmer's consent must satisfy Section 6. If it rests on a claimed research or statistical purpose, the Section 17(2)(b) exemption must not be allowed to become a laundering device for commercial model-training. And under Section 8(2), a fiduciary that engages a processor remains responsible for that processor's conduct. The State will bear the statutory liability for a downstream corporate misuse it enabled, but the injured farmer will bear the loss first.

What the Law Must Now Demand

The Act's substantive obligations are being brought into force in phases under the DPDP Rules, 2025. That sequencing is an opportunity: the architectural decisions of AgriStack are still being finalised, and compliance can be built in rather than retrofitted. The law should now insist on the following, none of which is negotiable.

AgriStack's custodians must be notified as Significant Data Fiduciaries under Section 10. The volume of data, the sensitivity of aggregate inference and the risk to the rights of the farmer population justify it. Notification would trigger a Data Protection Officer based in India, an independent data auditor and periodic impact assessments.

There must also be a dedicated, independent Data Protection Officer for rural registries. This cannot be a nodal officer drawn from the very department that processes the data. It must be a statutorily insulated functionary with a duty to report to the Data Protection Board and a rural-facing grievance mechanism consistent with Section 13.

Corporate data sharing must be frozen by statute. No State-held farmer dataset should reach a private entity until a specific statutory regime exists. Executive circulars are insufficient for rights of this magnitude, and the freeze should be secured by legislation or binding rules.

That regime should be built on agrarian consent sandboxes, codified in law. These should be supervised pilots in which consent is tested in the languages, literacy conditions and assisted-consent formats of actual farming communities, with the Data Protection Board supervising and publishing the results. Only a consent regime that survives this testing should be permitted to govern commercial sharing.

Benefits must be decoupled from enrolment. No cultivator should lose MSP or credit eligibility for declining to participate in secondary uses, since anything else collapses the "unconditional" requirement of Section 6. Finally, a rapid-correction channel for land-record disputes should run parallel to revenue-court litigation, so that a flagged error suspends adverse downstream decisions rather than awaiting a title adjudication.

The constitutional promise of Puttaswamy was that dignity and informational self-determination do not depend on literacy, wealth or the form of one's landholding. A farmer's data is not a raw material to be mined first and regulated later. If AgriStack is to be a foundation for rural prosperity rather than a ledger of rural vulnerability, the law must reach it before the algorithms do.

Author Sarvesh JP is Project Associate at Bay Of Bengal Programme Inter-Governmental Organisation (BOBP-IGO), Chennai & Deepanjana Saha is a Ph.D. Scholar at Ashoka Trust for Research in Ecology and the Environment (ATREE), Bengaluru. Views are personal.