10,000 Flights a Day: The Liability Stack a Drone Network Needs, Not a Drone Operator

On 24 August 2026, TechCrunch reported that Bengaluru-based Airbound had raised a $37 million Series A led by Greenoaks, with participation from DoorDash, Lachy Groom, Lightspeed and Humba Ventures. DroneLife carried the round on 25 August 2026 alongside the more consequential news for anyone underwriting Indian drone risk: Airbound has signed a commercial drone delivery agreement with the state of Andhra Pradesh for an aerial network across three districts, scaling to 10,000 flights per day and serving retail, e-commerce and healthcare deliveries.

The scale figure is the story. DroneLife and Business Standard both reported on 25 August 2026 that Airbound has flown more than 13,000 autonomous missions to date and raised close to $50 million since launching in March 2023. Thirteen thousand missions is a respectable demonstration record. It is also about a day and a quarter of flying at the target density.

Indian drone insurance has been priced against demonstration-scale operations. By early 2026 the DGCA had approved three commercial BVLOS corridors in the country, under the Drone Rules, 2021: Ladakh for minerals survey, Telangana for pharmaceutical delivery and Andhra Pradesh for coastal monitoring. Each has a defined purpose and a countable number of sorties. A single network flying 10,000 sorties a day is 3.65 million flights a year, a volume those purpose-limited corridor approvals were never framed around. That is a change in kind, not degree, and it lands on a liability framework built for the smaller thing.

Indian drone programmes placed to date are written per drone. A small drone in the 2 to 25 kg class typically carries a third-party liability sum of INR 1 crore to INR 5 crore, priced per airframe per year. At a few hundred sorties a month, the per-occurrence limit and the practical annual exposure are close to the same number. At 10,000 flights a day they separate, for three reasons:

1. Frequency alone guarantees repeat claims. Even at a loss rate low enough to be commercially viable, 3.65 million annual movements produce several third-party events a year as arithmetic rather than misfortune. The number that matters is the annual aggregate.
2. The events correlate. A network runs one airframe type, one autonomy stack and one maintenance regime across every district. A software regression or a battery lot defect does not produce one claim; it produces a cluster inside a short window, sometimes before the fleet can be grounded.
3. Ground exposure follows the route network, not the aircraft. Per-occurrence severity is set by the worst point on the route map, and a network delivering retail and healthcare across three districts crosses that point several hundred times a week.

The programme therefore needs a per-occurrence limit sitting under a separately negotiated annual aggregate, the aggregate sized against the correlated-cluster scenario rather than the single-incident one. Indian drone fleet policies commonly set aggregates at three to six times the per-occurrence limit, a fair convention for thirty airframes flying a few thousand sorties a year and a thin one at this density.

The Drone Rules, 2021 remain the operative framework for the Andhra Pradesh network. They handle third-party liability by importing the compensation machinery of the Motor Vehicles Act, 1988, applied to unmanned aircraft with the necessary modifications, rather than writing an aviation-grade liability schedule of their own.

That borrowing was a reasonable shortcut in 2021. It assumes the shape of a road accident: one vehicle, one incident, one identifiable operator, a claimant who can point at the machine that caused the harm. Those assumptions hold for a survey contract or a pharmacy shuttle. Neither holds cleanly for a network:

• Attribution. When a parcel or a component comes down on a vehicle in a market street, establishing which of the day's several thousand autonomous sorties it came from is a data-forensics exercise rather than an eyewitness matter. Flight-log integrity becomes a coverage issue, because an operator who cannot attribute an event cannot prove the flight was in permitted scope.
• Aggregation of claimants. One failure over a crowded location can produce many claimants from a single event, which a per-occurrence limit built on a single-claimant assumption will not answer.

The framework is also moving. In September 2025 the Ministry of Civil Aviation released the draft Civil Drone (Promotion and Regulation) Bill 2025 for public consultation, proposed to replace the Drone Rules 2021 entirely. A network built to a multi-year schedule will likely spend part of its life under the successor statute, so wordings that hard-code compliance to the Drone Rules 2021 by name, rather than to the civil aviation regulation in force, will need endorsement. Raise it at placement.

Hull insurance for a demonstration operator is a schedule of airframes with a per-unit insured value and a percentage deductible. For a network it is an operations-finance question.

The values are individually small and collectively material. Delivery airframes in this class are typically insured in the INR 8 lakh to INR 50 lakh range each, with hull premium around 6 to 12 percent of insured value annually. Tolerable on fifteen airframes, that rate becomes one of the larger operating cost lines across the fleet implied by 10,000 daily flights, and worth negotiating on loss data rather than accepting as market convention.

Deductible stacking is the real risk. A percentage deductible per airframe is harmless when incidents are isolated. A hailstorm across one district, a charging fault at a hub, or a firmware rollback affecting a batch of aircraft produces several simultaneous hull losses, each carrying its own retention. Negotiate an aggregate deductible cap per event and per policy year, or the operator self-insures a cluster loss at the moment its cash is under pressure from the grounding.

Fleet mechanics belong in the wording from the start. A network adds and retires airframes continuously. Run the programme on a declaration basis with periodic adjustment rather than an endorsement per airframe, and specify that a new aircraft is covered from the moment it enters service and is recorded in the operator's register, not from the date the insurer processes the addition. Charging infrastructure, launch pads, ground control stations and spares inventory are property exposures rather than hull, and belong in a property programme with its own limits.

The Airbound agreement covers retail, e-commerce and healthcare deliveries. Those categories have different loss profiles, and running them under one payload sublimit is the most common structural error at this scale.

Retail and e-commerce parcels are low value and high frequency, and the loss that bites is operational: a failure rate that breaches a service-level term. The question is whether the operator has assumed contractual liability to the platform for consignment loss, and whether that assumed liability is insurable under the placed wording. Contractual liability is frequently excluded or narrowly covered. Read the clause before signing the delivery contract.

Healthcare consignments invert the profile. Individual value is high, goods are often temperature-sensitive, and consequential loss can exceed goods value widely. Cargo and transit cover here should sit on a declared basis with a per-consignment limit reflecting the network's actual high-water mark, and cold-chain conditions written so a documented excursion is a covered peril rather than a breach of warranty.

The shipper interface decides who claims. Some consignments are already insured by the shipper under an open cover. Where the operator also carries payload liability, the same loss sits under two policies while the insured waits for the insurers to settle contribution and subrogation. Fix it in the delivery agreement: name which party insures the goods, name the other as loss payee, and record one first-response policy per class.

The delivery agreement answers this question, and answers it before the insurance programme is designed.

Three parties are in the frame for a ground-damage event. The operator flies the aircraft and holds the DGCA standing. The state commissioned the network and carries the political consequence of a public incident. The shipper, whether a retail platform, a marketplace or a health department, owns the goods and often the customer relationship at the delivery point.

Under ordinary tort principles and the Drone Rules framework, the operator is the party an injured third party will pursue, and both the operator entity and the remote pilot sit inside the DGCA accountability structure. That default is the starting assumption in every wording, and the commercial documents then move risk around it.

Indemnities running to the state and to shippers are the clause to read first. State delivery agreements and platform contracts both commonly require the operator to indemnify the counterparty against third-party claims arising from operations. That is assumed contractual liability, and the liability programme must respond to liability assumed under contract, not only liability at law. Stacked indemnities triggered by one ground event are the fastest route to a limit exhaustion nobody modelled.

Delivery to homes, shops and clinics also hands off at a point neither party controls, so the agreement should say when custody transfers: that moment separates payload liability from public liability.

The broker's job is to read the delivery and platform agreements alongside the policy wording and produce a one-page allocation showing, for ground injury, ground property damage, consignment loss and regulatory grounding, which party bears the loss and which policy responds. If that page cannot be written cleanly, the programme has a gap whatever the limits say.

The exposure easiest to underestimate at network scale is not a liability claim. It is the loss of the network itself. A serious incident triggers a DGCA review and, usually, suspension of the permission while the cause is established. That suspension attaches to the airframe type or the operating procedure rather than to the aircraft involved, so one event can stop all three districts while the service commitments keep running.

Three layers need separating. The hull and third-party sections address the physical loss and the liability claim. The other two are thinner:

1. Revenue loss during grounding, which business interruption cover for drone operations addresses only partially in the Indian market. Cover triggered by regulatory suspension rather than physical damage is uncommon and narrow where it exists.
2. The contractual consequence, meaning service-credit and penalty exposure under the state agreement and platform contracts. This is almost never insurable and needs capping in the contract itself.

Hold capital against both rather than assume a policy answers. Any regulatory-suspension trigger a broker can obtain, even sublimited with a long waiting period, is worth more than the same spend on a hull limit that will never be tested.

For the underlying compliance mechanics, weight classes, Digital Sky registration and Remote Pilot Certification, the DGCA drone operator insurance framework and the aviation liability breakdown for drone delivery startups cover ground this post assumes.

A network operator seeking terms at this density is asking underwriters to price something with no Indian loss history behind it. The submission has to do the work experience data normally does.

Flight and incident data at network granularity. Missions flown, hours by airframe type, and incident and near-miss counts with cause codes, expressed per thousand flights rather than per year. An operator reporting more than 13,000 autonomous missions has a real dataset, and normalising it turns it into an underwriting argument.

Route-level ground exposure mapping. Population density under the corridors, overflight of sensitive locations, altitude profiles and worst-credible-impact points. This justifies the per-occurrence limit and is the most persuasive document in the file.

The correlated-failure analysis. Fleet commonality by airframe, autonomy build, battery lot and maintenance provider, with the grounding and rollback procedure that caps how many aircraft one event reaches. Underwriters pricing an aggregate are pricing this document.

The contract stack and the regulatory file. The state delivery agreement, platform agreements and the indemnity clauses in each, with the allocation page described earlier. Alongside it, corridor and BVLOS permissions with their scope, Digital Sky registrations, Remote Pilot Certification currency, and a plan for transition if the draft Civil Drone (Promotion and Regulation) Bill 2025 replaces the Drone Rules 2021 mid-policy.

Build the programme in layers rather than as one policy: a primary third-party liability layer with a working aggregate, an excess layer reaching a per-occurrence limit that matches the route-map worst case, hull on a declaration basis with an aggregate deductible cap, payload split by consignment class, and property for hubs. Domestic capacity will write the primary; the excess layers for a network this size will likely need international aviation capacity through a specialty broker or a GIFT City IFSC route. Operators that get good terms at this scale treat the insurance file as a live operational record, updated as districts, airframes, pilots and corridors change, not a document assembled in the four weeks before renewal.