What do we observe about GST at the firm level?

by Ajay Shah and Atibhi Sharma.

The introduction of the Goods and Services Tax (GST) in 2017, was an important milestone in the evolution of Indian tax policy. By consolidating a fragmented web of central excise, state value added taxes (VATs), and local entry taxes, the reform promised a unified, destination-based consumption tax. This concept is generally termed VAT worldwide. The original document (Kelkar et. al. 2003) used the term GST in order to avoid confusion with the then-prevalent Central Value Added Tax (CENVAT).

At its core, the value proposition of a modern VAT is neutrality: it taxes only final consumption. This is achieved through the Input Tax Credit (ITC) mechanism, which allows businesses to offset taxes paid on inputs against their output tax liability. Taxes should flow seamlessly through the production chain without cascading or sticking to intermediate producers. In recent years, there has been considerable concept drift, and a loss of coherence in the GST as it is practiced today. Input tax credit - the beating heart of the GST - is blocked in many situations (Modi & Shah, 2026).

A key element of building knowledge in India about tax policy as it affects firms would be to study the firm data on the subject of the GST. To do this, we turn to the CMIE Prowess database and ask questions about what is going on. A remarkable fact about the CMIE firm data, however, is that almost nothing about the GST is observed. By its very design, the GST is invisible to firm accounting. Under Indian Accounting Standards (Ind AS) and Schedule III (which sets out the mandatory presentation format and notes for financial statements) rules under the Companies Act (2013), gross GST transaction flows are completely netted out of corporate Profit and Loss statements. Consequently, firm-level GST in India is functionally unobserved.

Foundations

GST is a tax upon value added where firms are not levied indirect taxes like earlier; final consumers are taxed. Rather than taxing gross turnover at each stage of production (which creates a cascading tax-on-tax), GST allows a registered business to deduct the tax paid on its inputs from the tax it collects on its outputs. In this system, the firm effectively acts as a pass-through tax collector, holding collected output tax as a statutory liability to be remitted to the government. The accounting identity governing a firm's net GST liability is:

Net GST Incidence = Output GST Payable - Input Tax Credit (ITC)

Here, `Output GST Payable' is the statutory tax collected by the firm on behalf of the government, `Input Tax Credit' (ITC) is the tax paid on inputs and capital goods, and `Net GST' is the cash deposited into the government treasury via the electronic cash ledger.

Consider three realistic scenarios:

Scenario 1: Clean manufacturing flow.

An FMCG manufacturer purchases raw agricultural products, packaging materials, and energy for Rs.1,00,000 plus 18% GST (ITC of Rs.18,000). The firm processes these inputs and sells packaged goods to distributors for Rs.1,50,000 plus 18% GST (Output GST of Rs.27,000). Net GST paid in cash equals Rs.27,000 - Rs.18,000 = Rs.9,000. The economic value added is Rs.50,000, and 18% of Rs.50,000 is exactly Rs.9,000. The tax system functions neutrally as a pure value-added tax.

Scenario 2: One ITC blockage example.

A textile or fertiliser firm purchases raw materials and logistics services for Rs.1,00,000 at 18% GST (ITC of Rs.18,000, split between Rs.10,000 on input goods and Rs.8,000 on input services). Statutory caps limit the output tax rate on finished fabric to 5%, yielding Output GST of Rs.10,000 on Rs.2,00,000 of sales. The firm accumulates unutilised ITC of Rs.8,000. Under Rule 89(5) of the CGST Rules (2017), cash refunds under S.54(3) of the CGST Act (2017) are restricted strictly to input goods, while refunds for ITC accumulated on services are legally barred. Now the effective GST paid by the firm is 10,000 - 10,000 + 8,000. This excess Rs.8,000 is not visible in the financial statements.

Scenario 3: Another ITC blockage example.

A steel firm spends Rs.50,00,000 on constructing a factory building, paying 18% GST (Rs.9,00,000). Separately, the firm spends Rs.10,00,000 on mandatory Corporate Social Responsibility (CSR) activities, paying Rs.1,80,000 in GST. Under S.17(5)(d) and S.17(5)(fa) of the CGST Act (2017), ITC is explicitly blocked for immovable property construction and CSR. Therefore there is an excess GST payment of Rs.10,80,000. Under Ind AS 16, the Rs.9,00,000 blocked GST is capitalized into Property, Plant and Equipment (PPE), while the CSR GST is absorbed into operating expenses. The blocked GST loses its tax identity entirely, cascading into product costs and depreciation schedules without appearing as tax.

What accounting standards say

Why do these substantial GST cash flows disappear from corporate financial reports? The answer lies in Indian Accounting Standards (Ind AS 115 and Ind AS 16) working in tandem with Schedule III of the Companies Act (2013).

Under Ind AS 115 (Revenue from Contracts with Customers), revenue is recognized only to the extent of economic benefits flowing to the enterprise. Because GST is an agency transaction, indirect tax flows are presented strictly as balance sheet items rather than Profit & Loss (P&L) line items:

• Sales revenue is reported net of Output GST.
• Operating expenses are reported net of Input GST, as input tax is treated as an asset (receivable).

Furthermore, because the balance sheet captures only a point-in-time snapshot of residual unadjusted receivables or payables at fiscal year-end (March 31), the reported amounts may not be large. A large enterprise may handle thousands of crores in gross GST transactions during the year, but because monthly output liabilities are continuously set off against input credits, for a typical non-financial firm, the year-end balance sheet item appears modest or negligible. This creates a misleading impression of a minor indirect tax incidence while concealing the true volume of tax transactions and unrecovered credits.

To see how these disclosure rules function in practice, we inspect audited financial statements across three representative major Indian enterprises:

1. Tata Steel

Prior to GST, excise duty was reported as a gross revenue component with a transparent deduction line on the face of the Profit and Loss statement.

Following GST adoption, this disclosure vanished completely as excise duty was no longer applicable. Revenue from operations is reported net of GST, with zero line items for indirect taxes on the face of the P&L.

2. Britannia Industries

In Britannia's post-GST annual report, we are able to observe the outstanding GST liabilities as of the end of the year combined with employee payroll taxes and TDS under Note 24 ("Other current liabilities").

Similarly, unutilized input tax credits at the end of the year is bundled under Note 17 ("Other current assets") under the generic heading "Balance with government authorities".

3. Shriram Finance

For financial firms, the accumulation of ITC is explained by Rule 38 of the CGST Rules, which lets NBFCs choose between a detailed calculation or simply keep 50% of eligible ITC each month, letting the rest lapse. Once a firm opts for either method in a year, it cannot switch back until the next year. Shriram Finance's separate disclosure of "GST credit receivable" follows the ICAI guidance note on Division III of Schedule III to the Companies Act, 2013 for NBFC that is required to comply with Ind AS, which provides for such credit to be shown under "other non-financial assets." Notably, this figure is cumulative since GST's introduction in 2017, not a single year's number. However, since this detailed disclosure is not mandatory, practice varies across NBFCs - some report it as a separate line item, while others club it under "Balances with government authorities".

Simultaneously, statutory tax liabilities are aggregated under "Statutory dues payable" in Note 29. This is a mix of TDS payable, GST payables, and other statutory dues.

What's observed in the CMIE database

Because corporate databases like CMIE Prowess compile financial data directly from audited annual reports, the database inherits the structural netting mandated by Ind AS.

When an empirical researcher queries CMIE Prowess for firm-level GST data, the active flow of GST is missing. Instead, Prowess only captures a set of residual variables:

• sa_sales: Captures sales revenue net of output GST.
• exp_gst (GST Expenses): This variable remains completely empty or missing for virtually all non-financial and manufacturing firms because GST is a balance-sheet pass-through.
• sa_indirect_taxes: Pre-2017, this field captured excise duty and sales tax. Post-2017, Prowess maps this field to "rates and taxes, net" in Other Expenses, capturing only non-creditable local taxes, municipal rates, or tax dispute write-offs.

To help fix intuition, here are numerical values seen in CMIE Prowess fields for the three firms in FY 2024-25:

CMIE Prowess Variable	Tata Steel	Britannia Industries	Shriram Finance
sa_sales (Sales Revenue)	Rs.1,32,516.66 Cr (Net of GST)	Rs.16,859.22 Cr (Net of GST)	Not applicable for a financial firm
exp_gst (GST Expenses)	Missing / Empty	Missing / Empty	Missing / Empty
sa_indirect_taxes (Indirect Taxes in P&L)	Very Low (Local rates & cesses)	Very Low (Excludes creditable tax)	Negligible

A Policy Proposal

Measuring GST at the firm level is essential for understanding the operational efficiency, tax incidence, and financial health of Indian firms. Without visibility into gross GST collections, input tax credits, and blocked credits, researchers, investors, and policymakers remain unable to evaluate how indirect taxes affect corporate behavior, capital allocation, and productivity in India. Therefore, achieving transparency requires an explicit and separate regulatory strategy focused directly on mandatory disclosures.

The ideal list of information required in a standardized P&L addendum disclosure (or mandatory Note to Accounts) includes:

• Gross Output GST collected on sales and turnover.
• Gross Input Tax Credit (ITC) claimed on procurements (distinguishing input goods, input services, and capital goods).
• Net GST deposited in cash via the electronic cash ledger (GSTR-3B).
• Total blocked ITC under Section 17(5) of the CGST Act (disaggregating amounts capitalized into Property, Plant and Equipment vs expensed into P&L).
• Pending GST refund claims under Section 54 of CGST Act along with a standardized ageing schedule.

To implement this, the requirement must be inserted into Schedule III to the Companies Act, 2013 overseen by the Ministry of Corporate Affairs (MCA).

This regulatory amendment carries three major benefits with minimal friction:

• Market transparency: Enables equity analysts, credit rating agencies, and lenders to assess working-capital drag from trapped ITC and true cost distortions from blocked credits under Section 17(5).
• Public economics research: Unlocks systematic firm-level empirical research on indirect taxation without relying on restricted tax return data.
• Near-zero compliance cost: Firms already compute and audit these exact figures monthly for their GST filings. Publishing a summary reconciliation note in annual reports imposes virtually zero incremental reporting cost.

Conclusion

At present, firm-level GST in India remains unobserved in public corporate databases. When researchers analyze fields in CMIE Prowess, they are not observing the true magnitude or incidence of GST, because P&L sales and expenditure fields exclude these indirect taxes by design. Attempting to estimate corporate tax elasticity, compliance, or indirect tax incidence from these database fields is fundamentally flawed. The complete, itemized record of GST payables and receivables - reported across monthly returns for outward sales (GSTR-1), auto-drafted input credits (GSTR-2B), and summary tax settlements (GSTR-3B) - remains locked inside the administrative database of the GSTN. Mandating addendum disclosures under Schedule III offers a simple, low-cost path to restoring financial transparency.

Bibliography

Kelkar, Vijay L., D. C. Gupta, Vineeta Rai, N. S. Sisodia, D. Swarup, and Ashok K. Lahiri. 2004. Report of the Task Force on Implementation of the Fiscal Responsibility and Budget Management Act, 2003. New Delhi: Ministry of Finance, Government of India. http://www.dea.gov.in/files/other_reports_documents/1.pdf.

Modi, Arbind, and Ajay Shah. 2025. "Input Tax Credit and refunds under GST in India: Conceptual and legal framework." Working Paper 44. XKDR Forum. December 2025. https://www.xkdr.org/paper/input-tax-credit-and-refunds-under-gst-in-india-conceptual-and-legal-framework.

The authors are researchers at XKDR Forum. We thank Mahesh Vyas, Arbind Modi, Megha Patnaik, Sanhita Sapatnekar and Susan Thomas for their comments and suggestions.

Supervising what you cannot inspect

by Maninder Singh Juneja and Renuka Sane.

In traditional financial regulation, supervisors are able to inspect the thing being regulated. For example, a scorecard by a bank was usually a short list of factors that one could comprehend. These factors could be traced to key documents within the bank, banks had a stated rationale for why they were being used, and there was some common sense in making weighted averages. The scorecard itself was static - the same weighted factors were applied over a period of time. Inspecting this process of constructing the scorecard and using it for management decisions was how trust was built. The AI world is different. Models are rented rather than built, change continuously, and behave probabilistically. They cannot be fully inspected even by the institution deploying them let alone by the regulator.

How should we then think of regulation?

One approach is to intensify the traditional approach where regulators demand more explainability, more documentation, and more validation. This will drive up the costs of compliance. But more importantly, this is ill suited to the new world where the technology changes rapidly, where the bank does not control the AI it uses, where there is no clear artefact that the bank can give the supervisor such as a model or a document (Board of Governors et al 2026). When regulators push traditional approaches, banks will respond by choosing AI models which are easy to document rather than the ones best for them, or defer AI deployments altogether. All these are unhappy consequences. What we need are policy makers who understand the live systems of the new world.

In this article, we analyse these emerging problems from first principles. We start from scratch, understand the landscape of market failure in the world of AI in banking, and think about how regulators can grapple with this world.

Our key idea is that AI systems resist replicability. Regulatory strategies that demand replicability will flounder or choke technology deployment. We suggest the regulatory standard applied should be AI deployments that are "supervisable" - the outcomes should be observable, they should be attributable to causes, and reversible by humans.

How AI is actually deployed

Before we get to the puzzles faced by regulators, we need to describe what AI in banking is. This categorisation is not unique to banking. Banking is simply where their consequences are regulated. Five modes span the range.

AI as Tool: AI augments a step, a search, a calculation, a first draft that the human controls. The human is cognitively engaged.

AI as Collaborator: The human and AI co-produce iteratively, and the human participates at every stage.

AI as Recommender: AI generates scores or options, and the human makes the decision. This is the classic human-in-the-loop, but credible only if the human can meaningfully interrogate the recommendation and not degenerate into cognitive surrender.

AI as Preparer: AI does the work and the human signs off. The approval here is closer to a check by a supervisor rather than a re-derivation.

AI as Autonomous executor (agentic AI): AI executes autonomously inside guardrails and the human monitors on exception. Emerging forms of agentic AI for banking run from single-task agents (a payment released, a service query resolved) to multi-step workflows and customer-facing agents that transact.

A central issue here is the true (de facto) role of the human. A reviewer who approves a thousand recommendations a day is not overseeing a model, the model is overseeing her. A feature of any deployment is the measured divergence between the model recommendation (and estimated uncertainty) vs. the human decision. Managers of banks will need to worry about relapses of human behaviour inside the organisation, a bit like how hospital managers worry about bad behaviour by doctors within their organisation.

What AI does to market failure in banking

We now shift gears to look at the standard knowledge on market failure in banking. Regulation may be justified when (and only when) there exist market failures which cannot self-correct fast enough, and there is adequate state capability in banking regulation to be able to correctly identify them and intervene. AI's distinctive feature is that it can cure several classic failures. Better default prediction reduces credit rationing, better fraud detection cuts deadweight loss, richer risk assessment lets banks serve customers they previously could not price, AI advisors help customers avoid some malpractices by the bank. But there are also some new problems that are anticipated.

Information asymmetry: This happens in banking at two levels: borrower-to-lender (adverse selection, hence credit rationing) and firm-to-consumer (product complexity, hence mis-selling). AI narrows the first through alternative data and may widen the second. On one hand, the customer armed with AI can see through many things proposed by the bank which are not in her best interest. But the consumer cannot observe why they were shown a product, offered a price, or steered toward a specific insurance plan. The sales process itself becomes more opaque. Personalised pricing approaches first-degree price discrimination, extracting consumer surplus. A single flawed model can mis-sell to millions simultaneously, converting isolated conduct failures of the pre-AI world into a big correlated event. And redress weakens when a denial comes from a model the firm itself cannot explain.

Systemic externalities from shared infrastructure: Each bank chooses its models, data sources and vendors to optimise its own performance. When multiple entities choose the same ones, the sector's exposures become correlated, which is a cost no individual bank prices in. We list the vulnerabilities below:

Correlated model risk: banks on similar models and the same foundation providers respond identically to an event. The regulator, at the system level, has to manage what happens when institutions move together, because no single entity has experience of such behaviour or of the impact synchronisation adds.

Third party concentration: one vendor's failure propagates everywhere at once. India has already run this experiment, when a ransomware attack on one shared technology provider knocked roughly three hundred cooperative and regional rural banks off the payments network.

Correlated cyber breach: shared stacks mean one exploited vulnerability is every institution's vulnerability. AI lowers the attacker's costs (automated vulnerability discovery, deepfake social engineering) and adds new attack surfaces (data poisoning, model inversion, prompt injection against agents that can move money).

Runs on banks at level 3. The bank runs of old were a queue on the pavement. Then we got to Silicon Valley Bank where over a weekend, customers took away money from the bank. Now we can be at level 3: autonomous agents managing customer cash can turn a shared signal into a self-reinforcing run at machine speed.

The various market failures listed above behave differently across the five modes of AI use. For example, when a human constructs the offer, opaque pricing can get contained. However, when an agent personalises autonomously at scale, this may become severe. Systemic correlation is moderate when AI advises and severe when fleets of similar agents act simultaneously. Any regulation that grades by model type alone, or by use case alone, misses half the object. The next question is the mode of regulation itself.

Regulatory strategy

Regulation can work in two ways. Process-based regulation is ex ante: it prescribes how the firm must operate, defines required controls, mandates oversight, validation standards, limits on autonomy. Outcome-based regulation is ex post: it prescribes ends, fair treatment, solvency, and judges results, leaving the choice of methods to the firm.

Outcome-based regulation is the efficient default. It is technology-neutral, so it does not ossify as methods change; it lets firms find the least cost route to compliance; and it does not require the regulator to understand the firm's production function better than the firm does. But it has important preconditions: the outcome must be observable and measurable; it must be attributable to the firm and, ideally, to the cause; and the harm must be reversible or compensable. Process regulation is the right departure from the default when those preconditions fail and where outcomes are unobservable, harm is catastrophic or irreversible, or damage manifests only systemically or with a lag.

Traditional doctrine treats these preconditions as given: examine the activity, choose the mode. This does not work for AI. Whether an AI deployment's outcomes are observable, attributable, and reversible is an engineering choice, which needs to be settled at design time. We suggest that regulation should mandate observability. This makes it possible to have an "outcome-based supervision" model. There are three ways to ensure observability.

1. Telemetry implies that the institution keeps a track of every decision such that the system records which version of itself it was using, what information it was given, and any time a person stepped in to overrule it. This makes outcomes attributable. That way if a certain group starts getting more (or less) approvals than before, the organisation can evaluate what caused the shift - was it the model, or the group itself. Such a capability is being mandated elsewhere in the world for similar use-cases (European Parliament and Council of the European Union, 2024). The regulator should also consider if it wants to set a minimum common telemetry standard.

2. Boundaries and rollback include putting caps on what the system is allowed to do, rolling out new updates to just a small number of cases first (instead of everyone at once), and having a tested plan for switching back to the older version if something goes wrong. This ensures that if a bad update slips through, it only affects a small slice of decisions.

3. Probes make bias observable. One way is "paired testing": one submits two applications that are exactly the same except for details that hint at things like someone's race or gender, and see if they get treated differently. One can also compare approval rates against the company's own normal levels. Together, these checks can catch an unfair credit model in just a few weeks, instead of waiting years to see who actually pays back their loans. One can also check rejected applicants against credit-bureau data to see which ones got approved by someone else, and how they fared.

These are similar to the idea of decision receipts that record which rules were applied to which facts and in what sequence for every decision made by a government or public system (Srivastava, 2026).

Over time, the supervisor should also build a repertoire of its own test cases drawn from incidents, complaints and examinations across the system and run it against every material AI deployment, much as stress-test scenarios are run against every balance sheet today. This will ensure that what surfaces in one institution becomes a probe for all others. It thus allows the regulator to set its own observability layer.

Process regulation is then reserved for the harms that are systemic, correlated, or irreversible at machine speed. For such events there need to be protections such as circuit breakers that halt things before they cause damage, limits on how much the systems can do on their own, model diversity so they don't all fail the same way, and rehearsed back-up plans.

What follows for the supervisor, the board, and the customer

For the supervisor: The unit of examination shifts from the model to the deployment, and the examiner's question shifts from "show me the validation report" to "show me the behaviour": what boundaries were set, what exceptions were thrown, what overrides were exercised, how far the system drifted from its baseline. Supervisors will also need to find the intellectual clarity to avoid a wide variety of extraneous claims about regulation of AI, e.g. the push for economic nationalism which has nothing to do with market failure.

For the board: A board cannot certify systems it cannot inspect or understand. Its role is to govern the framework which includes the limits on what the system is allowed to do, and making the rules for when a decision must be escalated to a human. The board then needs to continuously check the exceptions and overall performance.

Exception-handling should also be written into policy. If a problem stays unresolved beyond a defined size or time limit, it automatically gets escalated to the board. Internal auditors should double-check that the numbers are real. The board sticks to this audited framework and does not inspect the system directly.

If a board is asked to approve something they can't understand, they will default to saying no. But if you let them govern the limits and the exceptions instead, they can say yes.

The customer: The customer becomes part of the supervisory architecture. The widening firm-to-consumer asymmetry has a structural corrective the pre-AI world lacked: the customer now has AI too. Mainstream assistants abroad have begun connecting directly to users' accounts. If product terms (rates, fees, eligibility criteria) are mandated to be structured and machine-readable, the customer's own AI does the comparing, the explaining, and the policing of mis-selling, continuously and at zero supervisory cost. The redress channel weakened by opacity is restored the same way: an adverse decision should carry its reason to the customer, and what would have had to be different for the answer to change. This restores the ability to contest. The same asymmetry that AI widened, AI-equipped customers can close, but only if regulation hands them the data.

An example

Consider a debt-collection example. When borrowers fall behind on payments, the bank has to decide how to chase each overdue account. Contact methods differ in cost. Automated SMS and IVR (the automated phone system - "press 1 to pay") are cheap; having an employee actually phone the customer is expensive. So the bank builds a model that allocates accounts: cheap automated nudges for most, and the costly human call reserved for the accounts where the model predicts that talking to a person will actually "cure" the account (get it back to paying). Human calling works better but it costs more. The model is rationing an expensive resource to where it thinks it'll pay off.

With AI, the collection systems would retrain continuously or run reinforcement-style optimisation against a live reward (cure rate per rupee spent). The model would be moving on its own, faster than the review cycle, toward a target that is a proxy for what the bank may actually want. Under the conventional "inspect-the-artifact" approach, the model is checked before deployment, signed off, and reviewed on a schedule (say quarterly or annually). However, if AI is an active optimiser, a quarterly review discovers deterioration only after it has touched thousands of accounts. It may drift toward a mis-specified target in a way no one can read off the model itself. Further, if the collections model is a shared vendor product, or several banks fine-tune the same foundation model on similar data, they all go bad the same way at the same time when borrower behaviour shifts. In contrast, in a "supervisability-built in by design" approach, a small slice of accounts is deliberately kept on the previous allocation method, running live alongside the new model. This allows comparison between the old and the new in real time.

Every decision should keep a record of a few things: which version of the AI was used, what action it chose, which group of customers it was dealing with, and how things turned out in the end. This is how it helps: Say the AI's overall success rate is going up, but one particular group is quietly getting fewer phone calls from real people and doing way worse because of it. With all that recorded, that gap shows up right as it's happening. You can compare it against what's going well elsewhere, trace it back to a specific version of the AI, and undo it that same day, because the older setup is still up and running as a backup.

The company should never have to dig into the AI's inner "thought process" to realize its behaviour has taken a bad turn. None of this replaces the normal testing that the firm would do anyway. The difference is that here, the ability to observe what's happening, pin down what caused it, and shut it down are all built right into the system from the start, instead of being things you have to go do to the AI afterward.

Conclusion

AI deployments sit uneasily in conventional mores of banking regulation. The intelligence is rented, the behaviour is probabilistic, and harms can move at machine speed. We suggest that the response should not be more inspection of what cannot be inspected, nor blind faith in outcomes that arrive too late. Efficient regulation will emerge from observability, attributability, and reversibility. Regulators should require banks to build these.

References

Board of Governors of the Federal Reserve System, Federal Deposit Insurance Corporation, and Office of the Comptroller of the Currency. "Supervisory Guidance on Model Risk Management." SR Letter 26-2. April 17, 2026. https://www.federalreserve.gov/supervisionreg/srletters/SR2602.pdf.

European Parliament and Council of the European Union. Regulation (EU) 2024/1689 (Artificial Intelligence Act), arts. 12, 26(6), and Annex III(5)(b). https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-12.

Srivastava, Manish. "Digital Governance Needs Decision Receipts." Episode 73 of Big Ideas. XKDR Forum, June 1, 2026. Podcast, video, 13:06. https://youtu.be/WFX4ITb9yok

Announcement

Position for researcher in public finance and urban governance

XKDR Forum is looking for a full-time researcher with qualifications in economics, finance, law, public policy, or an MBA to work on public finance, municipal finance, cities, urban governance, and public health finance.

About XKDR Forum

XKDR Forum is a Mumbai-based inter-disciplinary group of researchers working in the fields of regulation, courts, household and firm finance, climate change, health, public finance management and land. In these fields, the group engages in academic and policy oriented research, and advocacy.

As a researcher who joins us, you will be part of our team that builds knowledge on the working of government and its organisations from the lens of laws, regulations and institutional capacity. As a research associate at XKDR Forum, you will work on project deliverables under the supervision of a research lead. You will be expected to work in person at the office premises in Mumbai.

Eligibility

The eligibility requirements for this position are:

1. Qualifications in economics, finance, law, public policy, or an MBA.
2. At least one short-form and one long-form publication in reputed publications.
3. Quantitative skills are preferred.

Candidates with work experience of 3-5 years are preferred. You must be comfortable working in an inter-disciplinary research environment with people from varying backgrounds such as data-science, economics, public policy and law. You must be curious and passionate about research and must be willing to work on independent outputs as well as in teams.

The remuneration offered will be commensurate with your skill and experience and will be comparable with what is found in other research institutions.

Interested candidates are required to fill out this form by June 30th, 2026.

Author: Prashant Narang

Prashant Narang is a researcher at TrustBridge Rule of Law Foundation.

On this blog:

• When remedies become regulation: The Karnataka High Court's intervention in food licensing and street vending, 9 June 2026.
• What happens when arbitration deadlines are missed, 2 April 2026.
• From Statute to Zero-Cost: Section 31A and the Bombay High Court's Zero-Cost Culture, 18 November 2025.
• How Indians rank their rights: what 26 interviews tell us about Article 19 and property, 30 October 2025.
• Evaluating India's Customs Authority for Advance Rulings (CAAR) and charting a path for reform, 11 March 2025.
• The Blind Spot in Indian Arbitration: Fees, Power, and Structural Oversights, 27 February 2025.
• Executive discretion in regulating private schools in India: Evidence from Delhi, 19 February 2020.

When remedies become regulation: The Karnataka High Court's intervention in food licensing and street vending

by Prashant Narang, Aryan Pandey and Indira Unninayar.

I. When public health litigation expands into regulatory governance

On 19 September 2025, the Karnataka High Court delivered its decision in Karnataka Pradesh Hotel & Restaurants Association v. Union of India. The case began as a routine industry challenge to the Food Safety and Standards Act, 2006. The judgment oversteps statutory adjudication to engineer regulatory design. It answers a real public-health worry. But litigation like this rarely stays within the parties before the court. The Court encroached into the executive territory with no consideration of whether the state is actually capable of implementing what it now directs. Such directions tend to produce selective enforcement and compliance costs that fall hardest on those least able to bear them.

The petition arose from a 2012 directive on licensing enforcement. The judgment was delivered nearly a decade and a half later by which time, the regulatory landscape and the affected ecosystem had evolved substantially. Street vending, food delivery and the law on informal work had all changed and all bore directly on what the Court now ordered.

II. What the petition sought, and what the Court ultimately directed

Hotel and restaurant associations had challenged orders to enforce the FSS Act and its regulations. The trigger was a letter dated 13 March 2012 issued by the State Food Safety Commissioner, acting on the Union instructions, requiring all States to enforce the Food Safety and Standards Authority of India's (FSSAI) licensing and registration regime. Every Food Business Operator' ("FBOs") had to obtain a licence or registration as a condition for continuing their business.

The petitioners contended that this requirement was impractical and arbitrary, especially applied uniformly to establishments of vastly different scale and capacity. The burden, they said, fell hardest on smaller operators. They went further, asking the Court to strike down swathes of the Act and its regulations as unconstitutional.

The Court rejected these constitutional challenges in their entirety and upheld the validity of both the Act and the Regulations, noting that the Supreme Court had already affirmed the Act. It restated food safety as a public-health aim and accepted the State's claim that the rules rested on scientific and international standards.

It then issued two directions with implications beyond the immediate dispute.

1. It directed the Union Government to classify restaurants into small, medium, and large categories and to enact separate laws or frame separate guidelines for each, observing that reliance on turnover-based thresholds alone, was impractical and insufficiently responsive to differences in size and operational capacity.
2. The Court directed the State government to introduce health and safety rules specifically for street vendors and food trucks, and to establish a mechanism to ensure strict oversight of their implementation.

These directions are what give the judgment its broader regulatory significance.

III. Expanded prescriptions sans diagnosis risk over-regulation, arbitrary discretion, and regulatory incoherence.

A. New rules directed without a policy diagnosis -

The judgment's biggest gap is that it never finds that existing regulation has failed. Nor does it explain why new, vendor-specific rules are required, and whether existing processes for licensing, inspection, and enforcement have failed. It even concedes that the licensing rules already impose hygiene standards on every operator.

The FSS Act already establishes a comprehensive enforcement architecture. Section 30 vests primary responsibility in the State Commissioner of Food Safety, while Sections 36 and 38 operationalise enforcement through prescribed methods and designated officers at the district level within municipal and local jurisdictions.

The Court should have asked two questions: were existing standards inadequate, and had enforcement failed? However, the judgment neither raises nor answers these questions.

The reasoning moves from a general observation about the informality of street vending directly to remedial directions that materially reshape regulatory obligations. It does so without identifying any institutional deficiency that might have justified such an expansive remedy.

B. The Street Vendors Act framework was overlooked entirely -

The Court acts as if street vendors operate in a regulatory vacuum. They do not.

The Street Vendors (Protection of Livelihood and Regulation of Street Vending) Act, 2014 ("SVA") was specifically enacted to balance livelihoods against congestion, public health and urban order. It overrides inconsistent municipal laws and works through town vending committees ("TVCs"), surveys, and certificates of vending. The SVA is not merely a procedural architecture; it embodies a considered normative choice by Parliament, that street vendors are rights-holders, entitled to livelihood protection, meaningful participation through TVCs, and procedural safeguards before any restriction on their vending.

By directing new health and safety rules for vendors without engaging with this framework, the Court implicitly undoes that normative settlement. It treats vendors not as participants with protected rights but as subjects of fresh regulation – inverting the very premise of the statute Parliament enacted for them.

The result is regulatory incoherence and it is worth being specific about what that means in practice. Under the SVA, a vendor acquires a certificate of vending through a TVC process that must include vendor representation; this certificate is her legal entitlement to occupy a designated vending zone. Under the FSS Act, she must separately obtain a licence or registration from FSSAI, subject to turnover thresholds and hygiene standards. The Court's direction would now superimpose a third layer: vendor-specific health and safety rules with a fresh enforcement mechanism. Each of these three regimes carries its own authority, its own compliance requirements, and its own enforcement officer.

C. The Court's directions assume state capacity that does not exist -

As far back as 2020, only 47% of town vending committees had any vendor representation; seven states had not notified schemes under the SVA, and in four states no compliant TVC had been constituted at all (Narang et al., 2020).

The enforcement machinery under the FSS Act tells a similar story. As of 2021, there were only 2,531 Food Safety Officers nationally for roughly one crore street vendors, with vacancy rates between 33% and 90% across states (Mishra & Khattar, 2025). Between 2018 and 2021, fewer than 1% of food adulteration cases ended in conviction. None of this means enforcement has stopped. It means enforcement has changed. When an inspector cannot police everyone, he polices whomever he likes – and scarcity only raises the price of his goodwill.

Piling fresh directions onto this will not help; it will hurt. Pritchett, Woolcock and Andrews (2010) examined three well-funded reforms (schooling in India, budgeting in Mozambique, land titling in Cambodia) that all failed for one reason: each demanded transaction-intensive implementation, millions of scattered discretionary acts no centre can supervise. Street-food safety is the same kind of task. It is transaction-intensive (a crore of vendors, countless daily sales), discretionary (each inspector judges hygiene on the spot), high-stakes (a failed check can end a livelihood) and opaque (the encounter leaves no record). On all four counts, the very dimensions Kelkar and Shah (2022) name as the hardest for any state to master, it scores about as badly as a task can.

The sequencing is backwards, too. Early state-building, Kelkar and Shah argue, should begin with low-stakes, high-visibility tasks, short feedback loops, correctable errors – and reach for hard ones only once capacity exists. The order to keep "strict vigil" over vendors does the opposite: it escalates coercion before building the institutions that would restrain it.

This dynamic has become characteristic of the Indian regulatory ecosystem. Shah's account of the history of Indian finance documents a pattern of regulatory agencies consistently engaging in micro-management whilst lacking the state capacity to enforce their own frameworks.

High discretion combined with low capacity does not produce zero enforcement; it produces selective, rent-seeking enforcement. When inspectors are too few to visit every vendor, they must choose whom to visit and a shortage of inspectors does not dilute that discretionary power, it concentrates and rations it. The fewer the officers relative to a crore of vendors, the more valuable each discretionary decision becomes, and the higher the payment it can command.

As Rai and Shah (2015) observe, the Indian state is too often strong as in scary but not strong as in capable: it commands coercive reach without the institutional depth to convert that reach into governance outcomes. Ordering strict vigil onto a system with 90% officer vacancies in some states therefore does not produce better public-health outcomes; it produces more rent-seeking. Inspectors arrive not on a fixed schedule but whenever they are short of cash, and vague, subjective standards give them the pretext to do so (The Seen and the Unseen, Ep 18). The Court's directions thus simply widen the regulatory perimeter within which this behaviour can operate.

D. Cross-jurisdiction comparisons are persuasive only when capacity is comparable-

The judgment leans hard on foreign examples to justify a strong licensing and enforcement regime. It cites international norms to rebut the claim that the regime is impractical.

But it ignores the conditions that make those systems work. Licensing does not work in the abstract. It needs capacity, trained inspectors, predictable procedure and firm limits on discretion.

The judgment itself notes that regulators such as the United States Food and Drug Administration recognise wide variation in the size and capacity of food establishments, and that enforcement is typically carried out by local health authorities. These details matter. They determine whether regulation produces overall compliance or its very opposite by way of uneven and discretionary enforcement.

This is where the comparison breaks down. The FSLRC (2013) treats foreign models as inputs to adapt, warning against any bid to "mechanically transplant ideas from elsewhere". International standards inform; they do not, on their own, justify a domestic enforcement regime. The court inverted this. It used the FDA comparison as the justification itself, without asking whether the administrative architecture that makes those powers function exists here.

Pritchett, Woolcock and Andrews (2010) show why that architecture cannot simply be assumed to exist. As per them when governments copy institutional forms from higher-capacity settings, the laws, the agencies, the enforcement powers, without first building the administrative foundations that make those forms function, the result is the appearance of reform without its substance. It is, in their words, no reform at all.

The FDA comparison does not establish that India's enforcement regime should be intensified. It shows only that the FDA works within machinery that makes its powers function. Transplant the powers without that architecture and you import the coercion while leaving behind the restraint.

E. Non-parties bear the burden of directions issued without participation -

The High Court has not abided by one of the basic principles of natural justice, audi alteram partem – the 'right to be heard' before any orders are passed against a person, as it has not 'impleaded' and 'heard' street vendors and pliers of food trucks, before proceeding to pass directions concerning them. Yet it ordered the State to write new health-and-safety rules for them and to keep 'strict vigil' over them, without studying who they are or what they face.

The regulatory burden falls on informal workers operating under constrained economic conditions. The court treats informality as a regulatory gap to be closed, vendors operate outside the system, so the system must be extended to capture them. Shah (2026) inverts this reading. Where the state's enforcement is slow and unreliable, operating informally is not evasion of good rules but a rational adaptation to bad institutions. Vendors build workarounds precisely because formal compliance offers little protection and predictable harassment. The state then misreads the adaptation as defiance and tightens the rules, which raises the cost of formality further and entrenches the informality it set out to cure. A direction to bring a crore of vendors under "strict vigil" is the next turn of exactly this cycle.

IV. Food safety is a compelling goal, but cannot justify prescription without basis

The strongest defence of the Court's approach lies in the public interest at stake. Food safety directly impacts public health and the FSS Act itself emphasises risk management, consumer protection, and preventive regulation. The Court did not draft rules itself; it told the executive to. . Read this way, the judgment can perhaps be seen as an attempt to prompt more effective implementation of an existing legal framework.

However, that defence, has limited force if any, as the Court does not explain the reasons why such directions pertaining to street vendors and food trucks were required in the first place, and how the existing enforcement mechanisms under the FSS Act were inadequate. Without a demonstrated failure, intervention at the level of design has little to stand on.

The promise of later consultation cures nothing. Consultation after an order to make rules is not consultation about whether the rules are needed at all. Once the outcome is predetermined, the space for meaningful policy deliberation is confined to that predetermined outcome.

The Court unfortunately moved too quickly from concern to prescription, and in doing so, blurred the line between ensuring lawful administration and reshaping the regulatory architecture itself.

V. Conclusion: Prescriptions must stay focused and relevant

The judgment reflects a growing tendency: courts shifting from reviewing validity to supervising regulation, especially under the banner of public health or public interest. Such interventions may be well-intentioned. But good intentions do not substitute for institutional competence. In this case, the Court's directions go beyond correcting unlawful administration to enter the terrain of regulatory design, without any demonstrated failure of the existing framework and without hearing those most affected by the outcome.

This tendency is not confined to any single domain. As Jain and Reddy T (2025) observe, reform through judicial diktat characteristically bypasses public consultation on questions that carry complex second-order effects. The adversarial courtroom is not designed for the stakeholder deliberation that sound policymaking requires. When it substitutes for that process, the people most affected, here, street vendors and food truck operators, bear consequences that were never examined.

Lon Fuller, in The Forms and Limits of Adjudication (1978), offers a useful framework for understanding why. Fuller identified a class of problems he termed "polycentric", those where the disposition of any single issue carries implications for every other, such that pulling one strand "will distribute tensions after a complicated pattern throughout the web as a whole". In such contexts, he argued, adjudication becomes institutionally incapable, because the affected party's participation through proofs and reasoned arguments loses all meaning when no advocate "could possibly present to the tribunal the grounds that must be taken into account in the decision".

The Karnataka High Court's directions bear precisely this character. A judicial mandate to introduce new health and safety rules for street vendors does not resolve a discrete regulatory question, it simultaneously displaces an existing framework under the Street Vendors Act, imposes fresh compliance burdens on informal workers already operating at the economic margin, adds enforcement obligations to a system strained by Food Safety Officer vacancy rates and multiplies points of regulatory contact where discretion can be monetised. Each of these consequences shapes the others, and that interdependence is exactly what Fuller's framework identifies as lying beyond the proper limits of adjudication.

The cost is not only procedural. Compliance burdens imposed without the capacity to administer them do not produce better governance; they tax the everyday enterprise of people operating at the margin and dampen the very economic activity the state should want to encourage. As Shah (2026) puts it, this is the "effervescence of creativity and invention that a poor country cannot afford to extinguish."

The lesson is that remedial ambition must be matched by remedial discipline. Prescription without diagnosis, and supervision without capacity, do not produce better governance. They produce the illusion of it.

References

Bedi J. and Narang P., 2020. Progress Report 2020: Implementing the Street Vendors Act. Centre for Civil Society.

Mishra G. and Khattar J., 2025. FSS Act: Need for enforcement and accountability in India's food safety regime. Bar and Bench. 26 June 2025.

Pritchett L., Woolcock M. and Andrews M., 2010. Capability Traps? The Mechanisms of Persistent Implementation Failure. Center for Global Development.

Kelkar V. and Shah A., 2022. In Service of the Republic: The Art and Science of Economic Policy. Penguin Allen Lane.

Varma A. and Menon M., 2017. Restaurant Regulations in India. The Seen and the Unseen. 15 May 2017.

Financial Sector Legislative Reforms Commission, 2013. Report of the Financial Sector Legislative Reforms Commission. Ministry of Finance, Government of India. 22 March 2013.

Jain C. and Reddy T P., 2025. Why reform through judicial diktat is fraught with perils. Times of India. 8 November 2025.

Fuller L. and Winston K I., 1978. The Forms and Limits of Adjudication. Harvard Law Review, Vol. 92, No. 2.

Rai S. and Shah A., 2015. Going from strong as in scary to strong as in capable. The Leap Blog. 25 February 2015.

Shah A. and Varma A., 2026. Why Freedom Matters | Episode 10 | Everything is Everything. Everything is Everything. 1 September 2026.

Ahluwalia R. and Shah A., 2026. Why Firms Build Economies Ft. Ajay Shah | Growth is Good | Ep 25. Foundation for Economic Development. 27 March 2026.

Prashant Narang and Aryan Pandey are researchers at TrustBridge Rule of Law Foundation. Indira Unninayar is an Advocate-on-Record, Supreme Court of India.

Trust in the Era of the AI-Informed Customer

by Maninder Singh Juneja.

A patient grappling with a longstanding eye problem was diagnosed with MGD (Meibomian Gland Dysfunction) by an eye specialist. To make sense of the medical jargon, she photographed her meibography report, the gland scan, and uploaded it to an AI, which confirmed the diagnosis and the line of treatment. A few weeks later, wanting to avoid buying shades of cosmetics she already owned, she uploaded a photograph of her stock and of the items she was about to buy to the same AI. Instead of commenting on the shades, the AI told her the waterproof eyeliner she had used for years was blocking the pores of the gland, like wax in a drain, and making her dry eyes worse. She stopped, and this one change brought immediate relief. The AI connected across domains, unasked. Neither the physician nor the cosmetics counter had connected the two. The ophthalmologist saw the eye. The counter saw the product. AI saw the person. Neither was wrong. Each was trained to look at one domain. Every institution serves its own; AI serves the customer. This is not a failure of individuals but a structural shift with strategic implications. In this article we examine these implications.

Verification in markets

For as long as markets have existed, buyers have had to trust sellers, because there was no way to verify their word, or the effort and cost were too high. The economist George Akerlof spent many days puzzling over the nature of the vegetable retail market in Delhi. In 1970, he offered new insights into the market failure that arises when the buyer cannot verify what the seller knows. He later won the Nobel prize in economics for this work which helps us understand asymmetric information.

The market economy combats asymmetric information using substitutes, brands, professional licensing, statutory audits, regulators, and guarantees. For example, the customer trusts the bank's brand name instead of reading the fine print, or the doctor's medical degree instead of evaluating the diagnosis. Half a century of institutional architecture in financial services is built on this logic.

The use of AI has brought about a foundational shift in trust. AI has made verification in many situations quite feasible. Customers who once accepted substitutes for trust in institutions can now check a claim, compare alternatives, or challenge a recommendation on an inkling of doubt. The barriers to specialist access, cost, time, language, jargon, have all collapsed at once. The result is not an end of trust, but the pillars on which trust rests have changed.

How it impacts the brand

The brand is an informational shortcut, one of Akerlof's substitutes for verification. It stood in for the customer's inability to verify. The bank's or insurer's reputation stood in place of the customer's understanding of provider soundness, and the terms and conditions of the contract. AI changes that, but the picture is complicated, because the brand connotes two things at once: A promise of what we will do for you, and an aspiration of who you become or which tribe you join by choosing us.

The promise of what the brand will do is now more verifiable. The customer's AI checks every promise before purchase, searching the wider internet, cross-referencing user reviews and triangulating from multiple public sources. The bank that claims to be customer-first has its complaint-resolution record extracted from annual reports or instantly summarised from X. The insurer that promises easy claims has its claim-rejection ratio surfaced against peers. Promises that survive verification strengthen the brand. Promises that do not are revealed in seconds.

The aspiration is not in the product, it lives in the customer. People buy Apple products partly because Apple-ness signals something about themselves. Customers bank with a private bank not for any major service benefit but for who else banks there. People ride Royal Enfield partly because owning one says something no specification sheet captures. AI audits the promise. It cannot interpret the tribe.

The split deepens in the AI-to-AI world, where the customer's AI transacts with the institution's AI. The customer only experiences the outcomes, settlement speed, dispute rate, complaint-resolution time and median application-to-approval. It does not watch advertisements. In this perspective, brands need to invest in advertising that bolsters tribal loyalty, but advertising that is supposed to bolster the promise is now less important.

How it impacts labour

Like the brand, the professional is going to be hit by the AI wave. The professional of the old world was, by training, organised around the domain. The cardiologist gave the right answer to the heart in front of her, not to the medicine cabinet or the financial situation at home. The mutual fund agent recommended within his manufacturer's product set, not against the seven existing funds in the customer's portfolio. The AI can see the landscape comprehensively and the professional has to now compete with it.

In most domains the gap between the bottom and the top decile of professionals has been wide. AI compresses the gap from below. The advisor who tests his recommendation with AI before delivering it catches the portability clause he had not considered. The relationship manager who has the model argue the customer's case against his own pitch will close more often. Every word the professional says can now be cross verified; the smartest professionals will go up against 'unsophisticated consumers' with more respect.

In recent research, Brynjolfsson, Li and Raymond (2023) studied 5,179 customer-support agents at a Fortune 500 software firm and found that average productivity rose 14% with access to a generative AI assistant. Within this overall average, novice and low-skilled workers improved by 34%, while experienced and highly skilled workers showed minimal gain. By this reasoning, AI-powered unskilled labour will be tough competition against skilled practitioners.

The pattern is visible in India also. At one NBFC, the productivity of fresh-college LAP underwriters rose 40% with AI assistance, while experienced underwriters showed minimal gain. The bottom of the labour quality distribution comes closer to the top.

As the floor rises, the implications are twofold. First, the customer's worst experience disappears, and with it the customer's reason to switch providers. Second, the professional's competence becomes the table stakes. They have to now provide what AI cannot supply: Trust earned over time, judgement under ambiguity, the willingness to take a customer-friendly call when things are not going well with the business. These are the qualities that live on the aspiration side of the brand, and they are the real moats of professional competence.

When the asymmetry reverses

So far the shift has run one way, AI in the customer's hand against the institution. The same architecture runs the other way too. Institutions have always known what customers did. But AI interactions reveal something deeper, what the customer considered doing. The questions asked, the scenarios tested, the decisions abandoned. These are cognitive traces, and they sit closer to intent than anything an institution has had access to before. If applied to underwriting, pricing or customer acquisition, they create a new informational advantage that did not exist a year ago.

The trust consequence runs deeper than the privacy one. Behaviour is what the customer did once it was done. Intent is what she rehearsed before she was ready to be seen. A customer can absorb the knowledge that her behaviour was logged. When she learns the institution priced her, or declined her, on the strength of her question, the breach is of a different order. The first asymmetry was about information she did not have. This one is about information she did not know she was giving. This may lead the customer to stop being candid with the one tool that was working for her, because she now suspects it is also working for the firms. The independence this piece began with is the first thing she loses when the architecture turns around. We will have to face a new world of consumer protection complexity, going beyond the simpler questions of data privacy.

What then survives in trust? Not the part that rested on the customer's inability to check. What survives is what AI cannot manufacture. Judgement under ambiguity, the call no model will take responsibility for. A relationship proven over time, the banker who backed the customer through a bad cycle and was proved right. The human presence in a hard moment, the advisor who delivers difficult news with care. None of this can be read off a document, so none of it can be verified, and so none of it can be commoditised. The trust that survives is the trust that was never about information in the first place.

The institutions that Akerlof described were built on substitutes for verification. Those substitutes served a purpose. They filled a gap the customer could not fill herself. That gap is closing. What remains when the substitute is no longer needed is the thing the substitute was always standing in for. Genuine expertise, honestly applied, in the customer's interest. The institutions that had that all along have nothing to fear from the informed customer. The institutions that were selling the substitute will find, quietly and permanently, that the customer has stopped calling back.

The scarce asset is the question

If one risk is that the institution reads the customer's question, the other is that she asks a wrong question. While AI can reduce the information asymmetry, friction has not entirely disappeared. It rests on arriving at the right question to ask. Speed and convenience applied to the wrong question produce a confident wrong answer faster. The customer who compares home loans on interest rate alone misses the prepayment clause. The customer who has already decided to switch insurers asks questions that confirm the decision. The informed customer is powerful. The misinformed customer with AI is powerfully wrong.

The question itself is the unclaimed opportunity. No bank will build a question set that surfaces its own prepayment clause weaknesses. No insurer will build a question set that exposes its own claim-rejection ratio. This needs independent actors: non-profits, researchers, consumer bodies, who can make a GitHub for the questions consumers should be asking financial institutions.

What the Boardroom should debate

Every earlier wave of technology was institution first. Computing, internet, mobile was adopted and absorbed by the organisation and then passed onto the customer on their terms. AI exploded in consumers' hand, 100 million users in two months. The Board is now governing businesses where customers will know as much if not more than the organisation.

The first institutional response to AI has been operational, adding chatbots, analytical tools, dashboards, voice bots, service automation, that are good cost-saving initiatives. The strategic question is the viability of the business model itself.

What happens when the customer arrives informed? Which elements of the value proposition survive verification? Which revenue streams depend on customer ignorance or high search costs? Which promises would survive an AI audit? Which parts of the sales process assume an information asymmetry that no longer exists? These are not technology questions for the CTO. They are business-model questions for the Board.

The institutions that emerge stronger will not be those that adopted AI fastest. They will be those whose value remains after verification becomes cheap.

References

Akerlof, George A. (1970). "The Market for 'Lemons': Quality Uncertainty and the Market Mechanism". Quarterly Journal of Economics, 84(3), 488-500. https://www.jstor.org/stable/1879431

Brynjolfsson, Erik, Danielle Li and Lindsey R. Raymond (2023). "Generative AI at Work". NBER Working Paper 31161. https://www.nber.org/papers/w31161

Reuters / Similarweb (2023). ChatGPT user-growth figures. https://www.reuters.com/technology/chatgpt-sets-record-fastest-growing-user-base-analyst-note-2023-02-01/

Reserve Bank of India. Annual Report of the Ombudsman Scheme. https://www.rbi.org.in/Scripts/AnnualReportPublications.aspx

Maninder Singh Juneja is a partner at True North. He serves on the boards of Pine Labs, Nivara Home Finance and Integrace, and has previously served on the boards of Niva Bupa Health Insurance, Federal Bank Financial Services and HomeFirst Finance. The author thanks participants at a talk at XKDR Forum for myriad good ideas, and Ajay Shah, Renuka Sane and Aditi Mascarenhas for comments on earlier drafts.

Chinks in market efficiency: A Melody story

by Ajay Shah and Atibhi Sharma.

The main paradigm in finance is the efficient market hypothesis which suggests that as soon as there is new information, the news is rapidly incorporated in the price. Markets are generally quite efficient, it's hard to find opportunities for supernormal returns.

But there are chinks in the armour. There are clear examples, worldwide, where market prices have been clearly wrong.

In this article, we show a recent Indian story which should be added into the Mistakes of Markets catalog. The share price of Parle Industries soared following the video of Italian prime minister Giorgia Meloni receiving a packet of Parle Products' Melody, gifted by Indian prime minister Narendra Modi. Parle Products, the maker of Melody, is an unlisted private entity. We also show some interesting global examples and offer some thoughts on understanding finance.

The "Signal" Ticker Confusion (2021)

In January 2021, after WhatsApp changed its privacy policy, Elon Musk tweeted 'Use Signal', referring to the free, open-source, and heavily encrypted messaging app. This pushed noise traders to purchase shares of Signal Advance (SIGL). The actual Signal app is managed by the non-profit Signal Foundation and is not publicly traded. Signal Advance, on the other hand, is a tiny medical detection devices manufacturer in Texas.

SIGL was an illiquid penny stock. It surged from \$0.60 to \$70.85 within three days—an 11,700% surge within three days. It took the markets six to nine months to return to pre-tweet levels.

Clubhouse Media Group (2021)

In the same month, on January 31, 2021, Elon Musk tweeted that he would be joining a room on Clubhouse, the then-viral invite-only audio chat application. The next day, shares of Clubhouse Media Group (CMGR), a penny stock entirely unrelated to the privately held Clubhouse app, surged. CMGR had recently rebranded itself as a marketing agency from a healthcare firm. The stock opened at \$10.85 on February 1, 2021, up from \$1.50 - \$2.00 from the previous days. It reached an all-time high of \$28.43 on February 15, 2021 (a near 1,400% increase from its pre-tweet baseline) and it took roughly four months to return to its pre-tweet level.

Zoom Video vs Zoom Technologies (2019-2020)

When Zoom Video Communications (ZM) filed for its IPO in 2019, investors rushed to buy shares, accidentally purchasing the shares of a defunct mobile phone parts manufacturer called Zoom Technologies (Ticker symbol: ZOOM), a penny stock trading at \$0.005. The stock surged 54,000% to around \$5. This confusion was not a one-off incident, as a similar trend was observed during the lockdown in the pandemic of 2020 when ZOOM went up 1,800% to \$20.90 until the US Securities and Exchange Commission physically intervened, suspending trading for Zoom Technologies for 10 business days and forced a ticker symbol change to ZTNO.

Bombay Oxygen Investments (April 2021)

During the second wave of COVID-19, when there was a shortage of medical oxygen in India, noise traders started looking for oxygen in the market. They landed on Bombay Oxygen Investments, a Non-Banking Financial Company that had exited the oxygen manufacturing business in August 2019 and had received RBI's registration certificate on December 31, 2019 for the same. The stock surged 131.3% in under 12 trading sessions, rising from Rs 11,025 on March 31 to Rs 25,500 intraday on April 20 and then, fell by ~ 50% to 12,700 in August of the same year.

L. G. Balakrishnan (2025)

In 2024, shares of LG Balakrishnan & Bros (a manufacturer of automotive chains) surged to a 52-week high because noise traders mistook it for the upcoming IPO of the consumer electronics giant LG Electronics India. The mispricing lasted only a day as investors realised their mistake.

Pan-Homophonic Events

Zhang et al. (2026) documented a new phenomenon of pan-homophonic events, where confusion between linguistics, trending keywords and stock names triggers sudden market volatility, specifically for a Chinese technology firm named Chuan-da-zhi-sheng. Since "Chuan-pu" is a loose and phonetic translation of Trump in Mandarin, noise traders phonetically interpreted the company's name to mean "Trump wins big". This unrelated traffic software stock essentially became a trading proxy for US political events. In 2016, when Donald Trump won the US presidential election, shares of the company surged 7.6% in a single day and then again in 2024, following major turning points in the US election cycle, the same stock repeatedly hit its maximum 10% daily upper circuit limit.

Ticker Confusion and the Limits of Arbitrage

Balashov and Nikiforov (2019) documented the systematic nature of these mix-ups. Investigating 254 pairs of stocks, they found that erroneous trades account for roughly 5% of all trading turnover in the smaller "shadow" companies. A classic example is Ford Motor Company (Ticker: F). Investors systematically assume its ticker is 'FORD' - which is actually the trading symbol for Forward Industries, a micro-cap manufacturer of carrying cases for medical devices. Similarly, a paper by Rashes (2001) studied the mass confusion between MCI Communications and a completely unrelated fund with the ticker MCIC. They found that while the co-movement between the two similarly-named stocks is statistically significant, it is not something that arbitrageurs can easily exploit because shorting an illiquid shadow company is costly.

Parle Industries

The Parle Industries episode is the latest entry in this ledger. Parle Industries is not a confectionary giant; it is a micro-cap company involved in infrastructure development, real estate, and paper waste recycling. Prior to the viral Modi-Meloni video, its market capitalization hovered around Rs.360 million, the price of a few apartments in Bombay. It was a penny stock with a share price of about Rs.5 and an average trading volume of 20,000 to 60,000 shares daily in the 6 months prior window.

• 9:30 AM (IST) - The Baseline: Market opened in India. Parle Industries opened at INR 4.95.
• 10:00 AM (IST) - The Event: Italian Prime Minister Giorgia Meloni uploaded a video to her official Instagram account and then her X account with the caption, "Thank you for the gift."
• 10:00 AM to 3:35 PM (IST) - Amplification: The video went viral. On X, the hashtag #Melodi trended, and the reel became the most viewed reel on PM Meloni's Instagram account. Simultaneously, Google Trends intraday data showed a spike: searches for "Parle share" and "Melody".
• 3:30 PM (IST) - Noise trading: By the market close, Parle Industries was locked into a 5% upper circuit closing at INR 5.25. It closed at a volume of 857,248 shares.

The stock hit the upper circuit for five consecutive trading sessions. BSE historical data shows that from May 21 through the end of the month, the stock's delivery percentage reached exactly 100%. The noise traders were taking delivery, believing it was worth holding this for multi day horizons.

Market Efficiency and the Role of Liquidity

These are examples of how prices can go wrong, exposing failures in market efficiency.

A better interpretation of market efficiency comes from focus on how clever people could exploit the mistakes of the noise traders. Here, we see the problems of financial market completeness (can you take an opposing trade?) and financial market liquidity (is the size of your winning trade big enough to matter?). A market inefficiency that is not exploitable will not be readily solved by the market. With small cap penny stocks there are no single stock derivatives that rational traders can short. In India, stock lending does not work so it is not possible to short sell and profit from the mistakes in the price. To the extent that better financial economic policy increases liquidity, it will, in turn, increase access to the correct tools for trading (single stock derivatives and stock lending). As a result, these problems will be diminished.

These problems are a reminder of the difficulties of small capitalisation stocks. Financial market trading works extremely well for large firms. We may perhaps apply a thumb rule in India of a minimum point of a market capitalisation of Rs.10 billion. But we do wrong to assume it is equally useful and equally effective for small firms. There is a certain social justice instinct in India, where we like to bring the glory of stock market listing to small firms, thinking that we are giving a helping hand to a weak firm. We need to be more cautious in the usefulness of this approach.

Financial markets are a remarkable information processing system. It is easy to disrespect the drama that is ceaselessly afoot. What is going on is that millions of clever people have been harnessed to constantly look at the world and make prices. These prices are the commanding heights of the economy and shape the resource allocation. Markets are not perfect, they are the best aggregation of what humans can figure out based on their self-interest.

Bibliography

Parle Industries' upper circuit to Signal's 5,100% surge: 5 mistaken stocks that triggered market frenzy, Surabhi Pandey, Moneycontrol, 20 May 2026.

#Melodi trends on X as PM Modi gifts Melody toffees to Italian PM Giorgia Meloni, DH Online, Deccan Herald, 20 May 2026.

190 Million And Counting: Meloni's Melody Moment With PM Modi Is Mega Viral, Abhinav Singh, NDTV, 21 May 2026.

Publicly Listed Zoom Video Communications: Traders Buying Zoom Technologies, Jonathan Garber, Markets Insider, 18 April 2019.

Want to Invest in the Zoom IPO? Make Sure You Buy ZM, Not ZOOM, Minda Zetlin, Inc.com, 18 April 2019.

Traders mistakenly invest in Clubhouse Media Group after Elon Musk tweets about a separate, private app with the same name, Natasha Dailey, Business Insider, 2 February 2021.

COVID-19: Bombay Oxygen shares up 256%; it doesn't even make oxygen, Business Today, 20 April 2021.

Massively Confused Investors Making Conspicuously Ignorant Choices (MCI-MCIC), Michael S. Rashes, The Journal of Finance, Vol. 56, No. 5, 2001.

How much do investors trade because of name/ticker confusion? Vadim S. Balashov and Andrei Nikiforov, Journal of Financial Markets, Vol. 46, 2019.

Quantifying the Linguistic Complexity of Pan-Homophonic Events in Stock Market Volatility Dynamics, Yunfan Zhang, Jingqian Tian, Yutong Zou, Xu Zhang, and Xiao Cai, Entropy vol. 28, no. 1, 12 January 2026.

Mistaken Identity: LG Balakrishnan Shares Surge as Investors Confuse It for LG Electronics India, Nishanth Vasudevan, Economic Times, 15 October 2025.

The authors are researchers at XKDR Forum. The authors would like to thank Susan Thomas, Amrita Agarwal, Aditi Mascarenhas and Jay Kulkarni for their valuable feedback and discussions on this piece.