Delhi's Draft Drone Policy: The Insurance Evidence to Ready Before It Lands

On 8 August 2026 the Delhi government's minister told reporters that the state's draft drone policy is ready and expected to receive approval shortly. The framing in every account of the announcement was economic: The Indian Express reported it as "Draft drone policy ready, aim to create 1,000 jobs", The Economic Times as a policy "likely to get nod soon" that "aims to create over 1000 jobs", and CNBC TV18 carried the same figure. The New Indian Express carried the same job-creation number on 9 August.

That is the whole of the public record. The draft text has not been released and no clause has been quoted. Anyone telling a Delhi operator what the policy will require of them is guessing.

The one structural signal worth reading is procedural. On 5 August 2026, days before the drone announcement, the Delhi government cleared a business reforms Bill introducing deemed approvals and a single-window system. New state policies are being fitted into that approval architecture. A deemed-approval regime is fast for the applicant and unforgiving about paperwork: if the clock runs and the file is short one document, the result is a rejection or a lapsed application, not a phone call. Operators who treat insurance evidence as a last-minute attachment will feel that design.

Before adding a state layer it is worth being precise about the layer already in force.

Commercial drone operation in India runs on the Drone Rules, 2021 (as amended), administered by DGCA through the Digital Sky platform. The parent statute changed recently: the Bharatiya Vayuyan Adhiniyam, 2024 replaced the Aircraft Act, 1934 as the governing aviation law, and the Drone Rules now sit under it. The statutory basis, the rule-making power and the penalty architecture all changed with it. We covered that transition in the Bharatiya Vayuyan Adhiniyam and aviation insurance post.

The compliance objects a commercial operator holds are:

• a Unique Identification Number (UIN) for each unmanned aircraft, registered on Digital Sky
• a Remote Pilot Certificate (RPC) for each pilot, issued through an authorised remote pilot training organisation
• a type certificate for the airframe, unless the model is exempt
• third-party liability insurance, mandatory for every category other than nano, with drone damage to third parties routed through the compensation machinery of the Motor Vehicles Act, 1988 applied to unmanned aircraft
• airspace compliance against the Digital Sky zone map: green (no permission), yellow (ATC permission), red (central government permission)

Note what the central regime does not do. It does not set a rupee sum insured. It requires that cover exist and leaves the amount to the operator and the counterparty. That silence is the gap a state policy or a state tender fills, and it fills it with a number. The mechanics of sizing that number by weight category are set out in our drone operator DGCA insurance guide.

Aviation and airspace are central subjects. A state cannot license a drone, issue a UIN, or authorise a flight in controlled airspace. What it can do is still substantial, because it controls the ground, the land, the police and its own procurement.

A state drone policy touches an operator through five levers:

1. Conditions on use of state-controlled ground and property. Take-off and landing sites on municipal land, rooftops of state buildings, corridors over state roads, and local authority permission for temporary flying sites.
2. Law and order permissions. Delhi Police issue prohibitory orders against sub-conventional aerial platforms, including drones, over the capital with some regularity, particularly around Republic Day, Independence Day, state visits and large public events. An operator with a valid UIN and a green-zone flight plan still cannot fly through a live prohibitory order.
3. Empanelment and procurement. Almost every state drone policy creates a panel of approved operators for government work: survey and mapping, land records, drainage and encroachment monitoring, disaster response, health logistics. Empanelment criteria are where an insurance clause with a hard number usually appears.
4. Incentives tied to compliance. Job-creation policies attach capital subsidy, skilling support or park allocation to conditions, and cover is a cheap condition for a state to impose and audit.
5. Local incident handling. When a drone hits a car, a hoarding or a person, the first responder is the local police station, not DGCA.

None of that duplicates the central rules. It sits above them. An operator can be fully DGCA-compliant and still be unable to fly a paid job in Delhi.

Because the Delhi draft is not public, the honest way to prepare is to work from what comparable Indian state drone policies and state tenders have asked for, and build evidence that satisfies the strictest version.

The four evidence requirements that recur

• A stated minimum sum insured for third-party liability, expressed per occurrence and sometimes in aggregate. Where the centre is silent, the state or the tendering department names a figure.
• The department named on the policy, as an additional insured, a co-insured or a loss payee. This is a wording change requiring an endorsement from the insurer, not a line on the certificate.
• A cross-liability or severability clause, so a claim by one named insured against another is not defeated by both appearing on the same policy.
• A certificate of insurance valid for the whole contract period, naming the specific airframes by UIN and the specific pilots by RPC where the fleet is small.

Two further requirements appear in higher-value survey and mapping work: professional indemnity cover for the deliverable, since a mispositioned survey output a department relies on is an error-and-omissions loss rather than physical damage, and cyber or data cover where imagery of government assets is stored and transmitted.

The realistic NCR profile is not a single-state operator. A survey, inspection or delivery business working out of Gurugram or Noida flies jobs in all three jurisdictions in the same week, with the same airframes and the same pilots.

Three jurisdictions means three evidence regimes in parallel, each with its own minimum sum insured, named-party convention, certificate validity period and renewal cycle. The failure mode is administrative rather than technical. Nobody crashes a drone because of a certificate. What happens is:

• a certificate lapses in one state while the master policy is live, and a job is stopped at the site
• the fleet grows by four airframes and the schedule filed with State A is updated while the one filed with State B is not, so the aircraft that flew the job is not on the state's list
• the master policy is written per drone rather than per fleet, so a substitution on the morning of a job puts an unscheduled aircraft in the air
• a per-project certificate names the wrong entity because the operator bids through one group company and flies through another

Each of these is survivable as a compliance nuisance. Each becomes serious the moment there is a loss, because the insurer's first question after an incident is whether the aircraft, the pilot and the operation were within the declared scope. An unscheduled airframe or an expired state certificate is the kind of discrepancy that turns a payable claim into a contested one, and the operator carries that gap. Programme-design mechanics for a fleet flying across corridors are set out in our drone logistics operator risk profile.

The structural fix

Move from per-drone policies to one fleet master policy with:

• a schedule of airframes by UIN and pilots by RPC, updated on a fixed monthly cycle rather than ad hoc
• automatic cover for newly acquired airframes for a defined period (commonly 30 to 60 days) pending addition to the schedule
• pre-agreed certificate issuance, so a per-state or per-project certificate is a same-day request rather than a fresh underwriting conversation
• a single third-party liability limit set at the highest of the three states' requirements, not the average

Underwriters do not load a multi-state operator because it flies more kilometres. They load it because the operator's control environment is harder to verify and the tail of its liability exposure is longer.

Four specific drivers:

Exposure heterogeneity. A single-state operator flying a repeated route over surveyed ground has a characterised third-party exposure. An operator taking work across three states over mixed urban density, highways, rail crossings and event sites presents a distribution the underwriter prices at the worst end rather than the mean.

Compliance discipline is priced at the weakest link. Insurers assess the operator, not the state. If certificate management, schedule hygiene and permission records are tight in two states and loose in the third, the whole underwriting file reads as loose. This is the most common reason a multi-state operator pays more than its loss record justifies.

Aggregation on a single event. Three states means more airframes and more simultaneous sorties. A weather event, a firmware fault across a model type or a grounding order can hit the whole fleet at once. Aggregate limits, not per-occurrence limits, are where this shows up in the quotation.

Claim-handling friction. A liability claim arising in Delhi is handled by a Delhi surveyor, under an FIR filed at a Delhi station, in a Delhi forum. Three jurisdictions means three sets of local counsel and a longer average claim life, which means higher reserves, which feed renewal pricing.

The consequence is that the multi-jurisdiction operator's premium tracks its documentation quality more than its geography. Operators presenting a clean schedule, a per-sortie permission log, an incident register and a populated safety management system get priced closer to the single-state operator. Those presenting a stack of per-drone policies and a spreadsheet do not. The underwriting factors are unpacked in our drone delivery aviation liability post.

None of the following depends on reading the draft, and all of it takes longer than the window between notification and the first empanelment round.

1. Reconcile your fleet register against Digital Sky. Every airframe flown should have a live UIN, and every UIN on the register should correspond to an airframe you still own and operate. Retired and sold airframes sitting on a schedule are a claims problem.
2. Reconcile pilots against RPC validity dates. Build a renewal calendar with a 60-day warning, not a 7-day one.
3. Pull the current liability wording and read the operating-territory and scheduled-aircraft clauses. If cover is written per drone or per project, start converting to a fleet master policy now.
4. Ask your insurer in writing what it takes to add a named insured and a cross-liability clause, and how fast a certificate can be issued. That turnaround time determines which tenders you can realistically bid.
5. Size the limit against the strictest requirement seen in any NCR tender in the last 18 months, not the lowest. Raising a limit at renewal is cheap. Raising it mid-year during a bid is not, and sometimes is not possible.
6. Build the permission log. One record per sortie: airspace zone and clearance reference, local police permission where applicable, landowner consent for the take-off site, pilot on duty, airframe UIN, weather at launch. This is the artefact that most improves both an underwriting submission and a claim defence.
7. Separate deliverable exposure from flying exposure. If clients rely on your survey outputs, the flying policy does not respond to a bad deliverable. Price professional indemnity separately.
8. Decide which legal entity bids and which flies, and make them the same one, or make the relationship explicit on the policy schedule.

Do items 3, 4 and 5 first. They are the only ones with an external dependency and a multi-week clock. Everything else is internal record-keeping you control.

When the Delhi policy is notified, four things in the text will determine how much work it creates for an operator, and they are worth reading first rather than reading the document front to back.

Whether it names an insurance figure at all. If it does, that number becomes the market floor for NCR work, including private contracts, because private counterparties copy state minimums into their own terms. If it does not, the numbers still arrive, through individual tenders instead.

Whether empanelment is centralised or department-wise. A single state panel is a one-time compliance cost. Department-wise empanelment across land records, drainage, health and disaster response means the same evidence pack four times over, each with its own validity period.

How it treats the deemed-approval clock. Given the business reforms Bill cleared on 5 August 2026, drone permissions may be fitted into a single-window flow with statutory timelines. That favours operators with complete files and penalises those who submit and then chase.

Whether it addresses the airspace reality of the capital. A policy targeting more than 1,000 jobs implies a meaningful volume of flying, which implies designated corridors, simplified clearance for specified use cases, or a concentration of activity in the outer districts. Which of those three it chooses tells an operator where to put its ground infrastructure.

Until the text is published, the useful posture is to fix the evidence chain. It is the same evidence chain every layer of this regime asks for, and it is the layer an operator actually controls.