Tamil Nadu Doubled Gig Accident Cover to Rs 10 Lakh: Re-Cutting the Aggregator GPA Programme

On 1 September 2026 Tamil Nadu doubled the accident cover available to gig workers in the state to Rs 10 lakh, and launched the Vettri Karangal scheme for construction workers alongside it. Reports on 3 and 4 September carried the same headline number and set it in a wider package of state welfare measures for construction and gig workers, including an overseas employment scheme.

Five days earlier the other end of the spectrum was on display. On 27 August, Uber, Eternal and Porter withdrew from the Karnataka Platform-Based Gig Workers Welfare Board, on the stated ground that they did not wish to sit on a statutory body created under a law they are challenging in the Karnataka High Court. Amazon India stayed on the board. Only 15 platform companies had registered with the Karnataka board at that point, collectively representing about seven lakh gig workers, with Delhivery, Namma Yatri and Yulu agreeing to join.

For an aggregator running the same delivery or mobility app in Chennai and Bengaluru, that is now two different social-security architectures sitting under one national insurance programme. Karnataka runs a contribution model: the platform pays a welfare fee and the state administers a fund, as set out in our note on the Karnataka welfare fee and what it forces aggregators to buy. Tamil Nadu is running a state benefit at a headline sum insured of Rs 10 lakh, which is higher than the group personal accident capital sum many platforms currently buy for their partner base. The programme design question is no longer national. It is state by state, and it lands mid-policy-year for most aggregators.

The doubling takes the Tamil Nadu line to Rs 10 lakh from the Rs 5 lakh level it replaces. Before anyone touches the commercial programme, three properties of a state benefit have to be written down and agreed with the client, because each of them breaks the assumption that the state number simply substitutes for the policy number.

• It is a scheme benefit, not an indemnity to the platform. The state pays a notified amount to a registered worker or the worker's family on the scheme's own terms. It does not respond to the aggregator as insured, and it does not answer a claim brought against the aggregator.
• Eligibility is set by the state, not by the platform's roster. Payment depends on the worker being registered under the state's machinery and meeting whatever activity or registration conditions the scheme sets. Our piece on eligibility thresholds and the aggregator cover gap sets out how quickly a threshold test strips people out of a nominally universal population.
• It is administered on state timelines. Scheme payouts follow verification and disbursement processes the aggregator does not control and cannot accelerate for a grieving family in week two.

Registration is where this becomes concrete. Five years in, the e-Shram portal records 31.89 crore unorganised-worker registrations, reported on 26 and 29 August 2026. That is a large base, and it is still a base of registrations, not a verified map of who is active on which platform in which state on the day of an accident. Underwriters will treat a state benefit as a potential recovery only to the extent the platform can prove registration for the specific claimant.

This is the question the CFO will ask within a week of the announcement, and the honest answer is that the state benefit changes the expected cost of the programme far less than the headline suggests.

Take a platform buying a Rs 5 lakh capital sum insured group personal accident on 40,000 active partners in Tamil Nadu. The tempting move is to cut the commercial sum insured to zero, or to a token amount, and let the state carry the death benefit. Three things argue against it:

1. Coverage scope. Group personal accident pays on permanent total disablement, permanent partial disablement scales and, where bought, temporary total disablement weekly benefits. A state accident benefit centred on death does not reach the partial-disablement claims that dominate two-wheeler delivery frequency.
2. Population mismatch. The state benefit covers registered workers in that state. The policy covers everyone on the platform's schedule. Cutting the policy to match the state benefit imports the state's eligibility test into the platform's own promise to its partners.
3. The commitment is contractual. Partner terms, onboarding decks and city marketing all promise a number. Reducing it because the state raised its own number is a communication the platform has to make to its own workforce, and to the Tamil Nadu board it now deals with.

The defensible middle path is to keep the commercial capital sum where it is, treat the state benefit as an additional recovery rather than a substitute, and take the improved expected net cost to underwriters as a rating argument instead of a cover cut. That is the same argument our group personal accident structuring guide for gig platforms makes for keeping the disablement scale intact when the death benefit is crowded.

Coordination of benefits is where a well-intentioned redesign quietly destroys cover. Personal accident is a benefit policy, so the contribution principle that applies to indemnity covers does not automatically apply. Insurers nonetheless write contribution and other-insurance language into group personal accident wordings, and some will now try to bring statutory scheme benefits inside it.

Three drafting positions are available, and they are not equivalent:

• Silent. The policy wording says nothing about state schemes. The benefit is payable in full irrespective of any state payout. Cleanest for the worker, most expensive for the insurer, and the position to hold if the platform is not cutting sums insured.
• Excess of state benefit. The policy responds only above the notified state amount. This looks cheap and behaves badly: if the worker was never registered, was registered in another state, or fails the scheme's eligibility test, the platform has bought a policy that starts at Rs 10 lakh and pays nothing on a claim where the state pays nothing either.
• Primary with a subrogation-style recovery. The policy pays first, in full, on its own terms, and the platform assigns or accounts for any state benefit subsequently received. This preserves speed of payment to the family and still gives the insurer the economic benefit of the state scheme.

Whichever position is taken, define it against the notified state benefit as a named amount rather than against "any government scheme". Open-ended language pulls in every future state notification, in every state, at whatever level it is set.

Gig work is mobile. A delivery partner who works in Hosur in April and Chennai in September moves between two regimes without changing employer, app or vehicle. A schedule keyed to state of residence at inception will misdescribe a meaningful slice of the population by month nine.

Build the schedule so that state is a data attribute, never a cover attribute:

1. One national insured-person definition. Insured persons are the platform's registered partners in India, defined by partner ID and activity test, with no state qualifier in the definition itself.
2. One national capital sum. Set the sum insured at the level the platform is willing to promise nationally. Do not write state-tiered sums insured into the schedule; state variation belongs in the coordination clause, not in the benefit.
3. State as a reported field, refreshed monthly. Report headcount by state of predominant activity in the monthly declaration so the insurer can see the mix move and so any state-scheme offset can be evidenced at claim.
4. Automatic additions and deletions. A monthly-declaration basis with automatic inclusion of new partners avoids the individual endorsement trail that breaks down at scale.
5. A named coordination endorsement. One endorsement, listing the state schemes recognised for coordination and the treatment (primary, excess or silent), amended by named addition when another state notifies a benefit.

The test to apply to the draft is simple. Take a partner who registered in Karnataka, moved to Tamil Nadu in month seven, and died in an accident in month nine while the Tamil Nadu registration was still pending. Read the schedule and the coordination clause and see whether the family is paid the full commercial benefit. If the answer depends on which state the accident happened in, the schedule has state variation buried in it and needs to be rewritten.

Two payers with different verification standards will produce two different answers on the same accident unless the sequence is agreed in advance. Set it at placement, and put it in the claims service standard, not in an email.

The sequence that works for the worker and holds up with the insurer runs in this order. The platform intimates the group personal accident claim within 24 hours of notification of the death or injury, on the strength of its own trip and duty records. The insurer settles on the policy's own evidence set, without waiting for the state to verify anything. The platform files the state scheme claim in parallel, with the family's consent and assistance, and records the outcome. Any state benefit received is then applied per the coordination clause agreed at placement.

Two practical points decide whether that works. First, the evidence pack differs: an insurer wants an FIR or police intimation, post-mortem report, death certificate and proof of engagement on the platform, while a state scheme will want registration proof under its own machinery. Collect both sets in one visit to the family rather than two. Second, an aggregator that has withdrawn from a state board, as Uber, Eternal and Porter did in Karnataka on 27 August 2026, still has workers who may be individually registered and eligible. Board participation and worker eligibility are separate questions, and the claims team should not assume that a corporate position on a statute closes off a benefit route for a family.

The reputational asymmetry matters here. Where the state pays first and the platform's own policy is still in verification at week six, the platform gets none of the credit for the cover it paid for.

Aggregators renewing in the next two quarters should treat the Tamil Nadu change as a rating and drafting exercise rather than a cover-reduction exercise. A workable brief to the market has five parts.

• State-split exposure data. Partner headcount and active-partner counts by state, monthly, for the expiring year, with the Tamil Nadu and Karnataka splits called out. Underwriters are pricing a national account with two known state offsets, and they should be given the split rather than a blended average.
• A single coordination position. Ask the market to quote the primary-and-recover structure as the base, with excess-of-state-benefit priced separately as an alternative so the CFO sees what the discount actually costs in uninsured band.
• Registration evidence capability. State what the platform can actually prove about worker registration, including e-Shram capture rates for its own base. A platform that can evidence registration gets credit for the state offset; one that cannot should not be asked to accept an excess structure priced as if it could.
• Unchanged disablement scales. Hold the permanent partial disablement scale and any temporary total disablement weekly benefit. These are the frequency layers, and no announced state accident benefit has been shown to replace them.
• A named-state coordination endorsement. One clause, amended by named addition as further states notify benefits, reviewed at each renewal instead of redrafted mid-term.

The direction of travel is more state-level benefit design, not less, given Tamil Nadu's September package extending welfare support to construction and gig workers and Karnataka's board continuing to sign up platforms. A programme built on a single national schedule with a named coordination clause absorbs the next notification with an endorsement. One built on state-tiered sums insured has to be rebuilt every time a state moves.