A Missile Hit a Steelworks Linked to an Indian Name. If Your Group Owns a Plant Abroad, Read the War Exclusion

On 16 August 2026, Russian missiles struck the Kryvyi Rih steelworks operated by ArcelorMittal in Ukraine. Moneycontrol reported the next day that a blast furnace and power units were damaged, and The Economic Times reported on 17 August that the company would restore the plant after a strike that killed two workers and injured 13. The Kyiv Independent placed the plant inside a wider set of Russian attacks across Ukraine that day, killing seven people and injuring 51. Financial Express and PSU Watch carried damage assessments on 17 August.

Indian coverage led with the Mittal connection. The plant belongs to a Luxembourg-domiciled group rather than to an Indian one, so the framing is loose, but the reflex it triggers is the right one, because the class of exposure is now genuinely Indian. Tata Steel runs plants in the Netherlands and the United Kingdom. Indian groups own assets across Africa, Central Asia and West Asia, and pharmaceutical, auto component and engineering companies have built or bought plants in jurisdictions carrying real conflict probability over a ten-year asset life.

Every one of those plants is insured under a property wording that excludes war on land absolutely, inside a programme built for fire, machinery breakdown and business interruption, not for a missile.

First, the exclusion. Standard property and industrial all risks wordings, in India and in almost every market an Indian group would place overseas cover in, exclude loss caused by war, invasion, act of foreign enemy, hostilities (whether war be declared or not), civil war, rebellion, insurrection and military or usurped power. That is a carve-out of the peril, not a sub-limit, and it is usually drafted with directly or indirectly language that catches consequential damage too. A fire started by a missile is not a fire loss under that wording, and the insurer will run proximate cause analysis to establish it.

Marine and aviation classes buy war risk back through separate war clauses, and that asymmetry is itself a source of programme gaps, covered here in the war exclusion gap audit. Property on land has no equivalent standard buyback. The market answer is a separate policy, not an endorsement.

Second, the pool. Indian risk managers reach instinctively for the terrorism pool administered by GIC Re, because for domestic assets it is the standard answer and it is cheap. It does not help here, for two independent reasons. The pool covers property located in India, so an overseas plant is outside its territorial scope whoever owns it. And the pool is a terrorism instrument: terrorism definitions turn on acts committed for political, religious or ideological ends by persons acting for an organisation. A missile fired by a state's armed forces at an industrial target in an interstate conflict is military action, and falls on the war side of that line.

The standalone market, written mainly out of London and the international specialty markets, sells hostile perils as a ladder. A buyer picks how far up to go, and each rung changes both the premium and the wording.

1. Strikes, riots and civil commotion (SRCC) and malicious damage. The bottom rung: strikers, locked-out workers, labour disturbances, riots, civil commotion and malicious acts. Most commonly bought, and often already written back into a property policy as an add-on.
2. Terrorism and sabotage. Acts for political, religious or ideological purposes by persons acting for an organisation, plus deliberate destruction of property.
3. Insurrection, revolution, rebellion, mutiny and coup d'etat. Organised armed opposition to a government, short of interstate conflict. This rung matters most for African and Central Asian assets.
4. Civil war. Armed conflict between factions within one state, with the character of war rather than rebellion.
5. War on land, including invasion and acts of foreign enemy. The top rung, and the only one that responds to the Kryvyi Rih fact pattern.

Most buyers stop at rung two or three because that is what the broker quoted. A policy sold as political violence in a slip may contain any of these combinations, and the only way to know which is to read the perils definition and the exclusions together. Where an SRCC and terrorism policy is in force, war on land is still excluded on its face.

Insurers do not treat these perils as one risk with one rate, and the sub-limit structure tells you what they fear.

SRCC is high frequency and low severity. A riot damages a boundary wall or halts a shift. Losses are frequent enough to model, contained and rarely total, so rates are lowest and limits often sit close to full declared value. Terrorism and sabotage are low frequency, high severity and non-correlated: one bomb damages one site, which is why that market can write meaningful per-location limits with reinsurance behind them.

War on land is the opposite of both. It is low frequency, extremely high severity and highly correlated. When war arrives in a country it arrives for every insured asset there at once, and keeps arriving for the duration of the conflict. An insurer writing war on land in one jurisdiction is writing a single bet, not a portfolio. Three consequences show up in every quote:

• Sub-limits fall sharply up the ladder. A policy commonly carries full value for SRCC, a substantial per-occurrence limit for terrorism, and a much smaller war on land sub-limit.
• Annual aggregates bind harder than per-occurrence limits. In a sustained conflict the second and third strikes matter more than the first, and an aggregate that reinstates only with insurer consent is the real cap.
• Cancellation provisions shorten. War perils are commonly written with short-notice cancellation, sometimes 48 hours, and with automatic termination on hostilities between named major powers.

The implication is timing. Terms harden and capacity withdraws as a threat becomes visible, so a plant in a jurisdiction with credible ten-year conflict probability is a buying decision now, not one to revisit when the news turns. A Construction News piece on 12 August 2026 made the same point about acts of war in construction contracts: the allocation of war risk has to be settled before the risk crystallises.

The Economic Times reported that ArcelorMittal would restore the plant. Restoring a damaged blast furnace and its power units in an active war zone is not a 12-month project on any realistic assumption. It depends on refractory contractors, imported equipment, grid capacity, and on the security situation permitting work at all.

The property programme on a large integrated steelworks typically carries a business interruption indemnity period of 24 or 36 months, the honest reinstatement horizon for a blast furnace. A standalone political violence policy on the same asset usually offers a shorter period, commonly 12 months and often less at the war-on-land rung. Underwriters set it that way for the reason they cut the sub-limits: the insured cannot rebuild while the shelling continues, so an open-ended loss-of-profits obligation would run indefinitely.

The result is a two-part shortfall on one physical event. The material damage recovery is capped by a war sub-limit well below the reinstatement cost of the plant, and the loss of profits recovery stops at the political violence indemnity period, even though the plant is still down and the property programme would have paid for another 12 or 24 months had the cause been fire.

Model the war scenario on the indemnity period in the political violence policy, not the property programme's. Run the number at 12 months of gross profit against the 24 or 36 months the board believes is insured. The difference is retained exposure, usually large enough to change the buying decision at renewal.

For an Indian conglomerate the coverage question is half the problem. The other half is how a standalone political violence policy interacts with the architecture already in place, and this is where placements go wrong quietly. Four questions have to be answered at placement.

A typical Indian group runs an India-domiciled master programme, a global property and BI policy issued to the parent, plus locally admitted policies in each jurisdiction to satisfy local regulation and give lenders a document they accept. The master responds on a difference in conditions and difference in limits basis. Those mechanics are set out in controlled master programmes with DIC and DIL, and the admitted question in admitted and non-admitted insurance for Indian multinationals.

A standalone political violence policy does not slot into that structure automatically:

1. Who is the first named insured? If the policy names the Indian parent and the damaged asset is held by a foreign subsidiary, the insurer will ask whose insurable interest is being indemnified. A parent's shareholding is not the same interest as an operating company's ownership of a blast furnace. Name the asset-owning entity and add the parent for its own interest.
2. Is it standalone or does it wrap the master? A policy written outside the master has its own limits, deductibles and claims process, and gets no benefit from DIC and DIL mechanics. That is usually right, but it should be a decision rather than an accident, and the deductibles should be checked against each other so a war loss does not fall between two retentions.
3. Is the cover admitted where the plant sits? Several jurisdictions require insurance on local assets to be placed with locally licensed insurers. A non-admitted policy issued to the Indian parent can be valid as a contract and still create a local regulatory and tax problem, and will not produce the certificate a local lender wants.
4. How do claims proceeds move? If the policy pays the Indian parent for damage to a foreign subsidiary's plant, the money has to travel back to fund the repair, raising exchange control, transfer pricing and tax questions far easier to settle before a loss. Where funds must reach the local entity, a locally issued policy or a loss payee arrangement naming it is cleaner.

Now the case most wordings handle worst. Assume the damaged plant supplies semi-finished steel or an intermediate to another plant in the same group. The receiving plant suffers no physical damage. It runs out of input and stops.

Standard business interruption cover requires damage by an insured peril to property at the insured premises, and the receiving plant fails that test. The contingent business interruption extension exists for this, but it is drafted for third-party suppliers, and three features routinely defeat an intra-group claim:

• The supplier must be named and is often required to be a third party. Some wordings define supplier so as to exclude entities under common ownership, on the logic that the group could have insured that plant's BI directly.
• The damage must be caused by a peril insured under the receiving plant's own policy. If the Indian plant's policy excludes war on land and the overseas plant was destroyed by a missile, the extension does not respond even where the supplier is named, because the trigger fails at the receiving policy.
• CBI sub-limits are small. A few crore of contingent sub-limit against a downstream loss in tens of crore is a rounding error on the exposure.

There is also a double-counting problem. If the overseas plant's political violence BI cover pays its loss of gross profit and the Indian plant separately claims CBI for the same output, insurers on both sides will argue about whether the group is recovering the same margin twice. Decide up front where the inter-company margin sits and insure it once, in a group BI schedule identifying the flows by volume and value.

The test to run before renewal. Take the largest single inter-company input flow in the group, assume the supplying plant is destroyed by an excluded war peril, and trace every policy that could respond. In most Indian group programmes, none of them do.

The Kryvyi Rih strike is useful because it is not an Indian loss. It lets a group run the exercise cold, with no claim in progress.

1. List every asset outside India by entity, location and declared value, including leased plants, joint ventures and stock at overseas sites.
2. Extract the war and terrorism clauses from every policy touching those assets: master, local admitted, marine and standalone. Note which use directly or indirectly formulations and where the definitions sit relative to each other.
3. Rank jurisdictions by conflict probability over the asset life, not the policy year. A ten-year furnace campaign runs against a policy that renews annually with cancellation rights attached.
4. Check where the cover stops on the ladder, then price the top rung and take the decision consciously.
5. Reconcile the indemnity periods and report the difference, in months of gross profit on the largest exposed asset, to the risk committee as a retained number.
6. Map inter-company dependency: which overseas plants feed Indian operations, at what volume, and whether any policy responds when the feeding plant is hit by an excluded peril.
7. Fix the entity and proceeds mechanics. Confirm the first named insured is the entity with the insurable interest, whether local admitted cover is legally required, and how a payment reaches the entity that has to spend it.

That exercise means reading war and political violence language clause by clause across a stack of insurer wordings. Sarvada gives commercial insurance brokers structured, searchable access to insurer policy wordings, so an exclusion map for a client's overseas asset programme can be built before a loss rather than during one. Request Access to see how a group's overseas plants hold up when the cause of loss is a state military act.