Under the Hood: How Defence Startups Really Protect IP

To listen to an audio recording of this article, click here.

Two Defence CEOs, One IP Question

Ask a room full of defence company founders how they think about intellectual property and you will most likely hear:

Patents are everything.
Patents are useless.
Trade secrets are safer.
NDAs don’t mean anything.

The answer, as it turns out, depends a lot on what the company is actually building and how exposed it becomes once development transitions into real‑world use.

That distinction came through clearly in two recent conversations I had with CEOs working on very different problems in the same broad ecosystem. Both are serious technologists. Both are operating in defence. And both have landed on very different IP strategies, intentionally.

Teleidoscope: Hide the Magic, Ship the Results

Teleidoscope builds real‑time tracking, sensing and targeting technology for integrated air‑defence and counter-drone systems. The company’s proprietary tracker continuously adapts to how a specific object looks, moves and changes even as conditions degrade: cluttered backgrounds, occlusion, scale changes, frame drops or momentary loss of visual detail. This matters because modern air‑defence and counter‑drone systems do not just need to see threats, they need to hand off reliable, high‑confidence tracks to downstream systems, whether that’s human operators, interceptors or autonomous responses. That ability to self‑diagnose and recover is why the company’s software has been dropped into real‑world air‑defence systems, including those protecting some of the most sensitive airspace in the United States.

On the surface, this looks like a classic hardware‑driven defence business.

But when you dig in, the real value is not the hardware at all. It is the software, specifically the algorithms that pull clean, high‑fidelity targeting data out of messy sensor inputs.

For CEO Matt Rabinovitch, that fact makes the IP decision surprisingly straightforward.

“When we give updates to our Air Force customer, they’re not getting the source code,” he told me. “They’re getting binaries. There’s not really visibility for any external party to see what we’re doing under the hood.”

That sentence alone explains most of Teleidoscope’s IP strategy. It does not patent its core technology. It keeps it secret.

Matt is blunt about why patents do not appeal much in this context. “If you can’t detect infringement and you don’t know what someone’s doing under the hood, why bother?”

It’s a fair question. Patents can only be enforced if you can tell when someone is copying you. In algorithm-heavy systems, that’s often easier said than done. Disclose too little and the patent is weak or can be invalidated. Disclose too much and you have just given your competitors a helpful starting point.

“It’s always a fine balance,” Matt said. “If you’re not giving up enough, it’s not really defensible. If you’re giving up too much, somebody can take it and change it just enough.”

So Teleidoscope leans into secrecy, architecturally and operationally. No cloud infrastructure. No source‑code distribution. Tight control over how software is delivered and updated. The IP protection is baked into how the product exists in the world, not just how it is described on paper.

The same thinking carries over into Teleidoscope’s dealings with government customers. Having developed its technology prior to entering defence procurement, the company is able to address ownership questions with unusual clarity.

“We can say, ‘We already developed all of this before working with the government, so we retain ownership,’” Matt explained.

It is the kind of distinction that tends to matter only when it suddenly does.

Xubin Aerospace: Stake Your Ground Early

Meanwhile, Xubin Aerospace is taking a very different approach to drone detection, one that does not rely on radar, GPS or heat signatures. Instead, its flagship system, Eagle Vision, looks at the air itself. Using standard electro‑optical and infrared sensors paired with AI‑driven computer vision, the system visualizes the tiny aerodynamic disturbances created by a drone’s propulsion.

All of the heavy lifting happens at the edge. The AI models run on local hardware, making the system usable in mobile or remote deployments without needing constant connectivity. Xubin has already demonstrated the technology at a mid‑readiness level, achieving kilometre‑scale detection using relatively primitive hardware, and is now pushing toward multi‑kilometre ranges with higher accuracy.

Not only can Eagle Vision tell the difference between a bird and a drone, but it can also distinguish between different classes of drones.

Here, secrecy alone is not enough.

“We have pending patent applications,” CEO Mostafa Najafiyazdi said. “We’re also protecting different form factors – two or three design variations.”

Why the emphasis on patents?

“For now, IP for us is a deterrence,” he said.

That word “deterrence” comes up a lot in defence, and not just in the geopolitical sense. Xubin isn’t patenting because it expects to license aggressively or sue competitors tomorrow. It is patenting because once your system is in the field, pretending secrecy alone will protect you is naïve. Xubin’s patent applications are meant to plant a flag early, before larger players start paying close attention. In sectors like defence, attention itself can be a liability if boundaries are not clearly marked in advance.

But the company has been careful to avoid the all‑too‑common impulse to patent everything. “The overall system and how it works is patented. But there are specific implementation details that we keep as trade secrets,” he said.

Performance KPIs, for example, are completely off limits. “Performance metrics are considered trade secrets,” Mostafa said. “First, for competitive reasons. Second, because they fall under government‑protected categories.”

And when it comes to demos, Xubin is famously tight‑lipped.

“We had investors who wanted to see the device,” he said. “Even under NDA, we said no. Not before a term sheet.”

In Mostafa’s view, NDAs are necessary, but they are not magic. “An NDA is just a formality,” he said. “It’s only as good as the person signing it.”

Because Xubin’s system is intended for operational deployment, testing and handling, some degree of visibility is unavoidable. The company has planned for that reality, using patents to define the outer limits of its technology while relying on trade secrets to protect its most sensitive elements. Access is managed carefully and granted only when there is a clear commercial or strategic rationale to do so.

Same Industry, Different Realities

It is tempting to ask which approach is “better.” But that’s the wrong question.

Teleidoscope is able to keep its most valuable technology largely out of view because its advantage lives in software algorithms that customers never need to inspect directly. Xubin, by contrast, is building a system that must eventually be deployed, tested and handled, which makes a degree of visibility unavoidable. Same legal tools. Very different exposure profiles.

What’s interesting is how intentional both strategies are. Neither CEO is following a template. Neither is collecting patents, or avoiding them, out of habit. In each case, the approach flows from how the technology is built, delivered and used in the field.

In practice, many IP missteps do not come from bad decisions so much as outdated strategies. A strategy that made sense at the prototype stage quietly stops fitting once customers, partners or governments get involved.

The companies that navigate this well are the ones that keep revisiting the same uncomfortable questions:

What actually needs to be disclosed?
What can competitors realistically see?
What will endure after novelty fades?

Aird & McBurney LP works with technology companies to design and implement IP strategies that reflect how their products are built, deployed and commercialized. Please contact the author or a member of the team if you have any questions or require assistance.