When Does an Off-the-Shelf Machine Become Military Equipment?

Is “dual-use” about the item, or about the use?

Both, with the weight on the second. Dual-use describes a technology, piece of equipment, or material with a genuine civilian market and direct military value, and in most cases the item itself is identical in both uses.

A five-axis machine tool cuts aerospace parts and medical implants in one plant and missile casings in another. Same machine, same operation; what differs is the consignee on the order.

This is what most people find counterintuitive at first: control lists are written in specifications, but enforcement asks where the goods will end up. The specification is a filter; the judgement turns on the use and the recipient.

So how is it actually decided whether a transaction is controlled?

Ask four questions in order: what is it, what is it for, who receives it, and where does it go? A transaction is clean only when all four have clean answers.

The first is the easy one, because export control lists carry explicit technical parameters you can look up. The second is where it gets harder: the stated end use is on the paperwork, and the paperwork was written by the buyer.

The third has a tool: an entity list names known high-risk parties so that names can be screened against it. But a list covers only parties already identified, and a company incorporated last month will not be on it.

The fourth is the hardest, because it includes transshipment: when goods are lawfully exported to country A and re-exported to country B, the exporter is fully compliant on paper, but the real destination is not the one on the form. That is why due diligence usually turns less on the document than on where the document stops making sense. When a small trading company orders equipment far beyond its own scale, the anomaly lies in the proportion, not the equipment.

Why do the lists always lag?

A list can only describe what is already understood. To control a technology, somebody must first identify its military value, write enforceable technical parameters, and make those parameters precise enough not to sweep in an entire civilian industry, and that sequence takes longer than it takes the technology to iterate.

Commercial drones are the clearest case. Most of their components come from the consumer electronics chain: flight controllers, camera modules, batteries, communications chips. None looks like a munition on its own, yet assembled they make a usable reconnaissance and strike platform. Controlling them would mean controlling consumer electronics.

So the practical answer is usually not to chase the list but to move the lever: what cannot be stopped at the border is handled through purchasing instead. Governments and critical-infrastructure operators decline to buy airframes and components from particular sources, shifting the risk decision from customs to procurement specifications. That is one of the clearest recent shifts in the defence industrial base, and it is why many countries’ drone policies read as industrial policy rather than control policy.

Who actually gets caught by this?

It is not the arms manufacturers; they have had compliance departments for decades. Those swept in are usually of three types: component suppliers, service providers, and maintenance firms.

Component suppliers struggle because they cannot see the end customer. A connector goes to a module maker, the module maker sells to a systems integrator, and who the integrator sells to is commercially confidential. The “knew or should have known” standard therefore lands on the least-informed link in the chain.

Services and maintenance are overlooked more often still. Under many control regimes, technical support, remote diagnostics, training, and software updates are treated like physical exports, because what moves is knowledge, and customs never sees it. Sending an engineer on a trip can be legally equivalent to shipping a machine.

The lesson is the same for all three: compliance turns less on whether the form was signed than on whether anyone actually looked at the part of the deal that does not add up.

What does this mean in practice for companies in Taiwan?

Taiwan’s industrial structure makes this immediate: a great many firms supply precision machinery, electronic components, materials, and modules, exactly the categories that dual-use lists cover most densely. You do not have to consider yourself part of a defence industry to fall inside the scope.

Three things are cheap and effective: turn the end-use statement from routine paperwork into a document someone actually reads; screen the recipient’s name and address against public lists, and repeat the check at every renewal rather than only at onboarding; and write resale and re-export restrictions into the contract so the file can at least show you asked.

Overreaction has a cost too. Refusing everything adjacent looks safe but outsources judgement to fear, and fear does not distinguish real risk from an unfamiliar customer. The line should be drawn by the four questions, not by instinct.