The AI data supply chain is global. That is mostly a good thing. Labelers in the Philippines, India, and Eastern Europe produce high-quality annotations at price points that make large-scale commercial ML economically viable. The global labor pool is why commercial autonomy exists.

It is also why commercial autonomy's supply chain doesn't work for defense. Capture, in particular, has a different set of constraints that make offshore sourcing a non-starter for most program offices.

Why Capture Is Different From Annotation

Annotation is a data transformation performed on data that already exists. The data can be transferred in a controlled, audited way to an accredited environment, and the annotation team operates on it under contract. If the original data is cleared for foreign-person access, the work is viable.

Capture is the act of creating the data in the first place. The crew is physically present at the collection site. They operate sensors tied to program requirements. They make in-the-moment decisions about framing, framing accuracy, and scenario execution. The data they produce is the data — there is no sanitization step between the sensor and the file.

Three consequences follow, and they all point to the same place.

Clearance And Citizenship Requirements

Most defense capture contracts require crew members to be U.S. persons, frequently with secret or higher clearances. This is not arbitrary. It flows directly from ITAR, from site access controls, and from the deemed-export considerations that come with operating on or near sensitive installations. An offshore crew doesn't clear the bar regardless of technical proficiency.

Operational Security And Intent

Crews on defense capture work know where they are, what they are capturing, and why. That knowledge is itself sensitive. An operator who executes a collection plan against a specific platform in a specific environment has seen things they should not discuss publicly. The hiring pipeline, the vetting, the ongoing relationship, and the nondisclosure posture all need to match the sensitivity of what the crew will be exposed to. That is only sustainable with a crew you control and can vet directly.

Responsiveness And Trust

Defense capture is rarely a one-shot event. It is a relationship. The program changes, the requirements shift, the platform evolves, and the capture plan has to keep up. A crew that can turn around a re-collection in 48 hours, take a weekend call about an emerging requirement, and show up at a short-notice range is worth substantially more than a cheaper crew that operates on a different continent and different time zone. Latency is a cost in this business, and geographic distance adds latency everywhere — scheduling, coordination, escalation, iteration.

What A Defensible U.S. Crew Looks Like

  • U.S. citizens, background-checked, with NDAs and clear records of the work they have been exposed to
  • Part 107 certified minimum, with COA/SGI familiarity for crews flying in regulated airspace
  • Trained on the specific payloads — not generic drone pilots, but operators who know the sensing stack cold
  • Geographically distributed across the continental U.S. so that capture in a specific environment doesn't require moving the entire crew
  • Under direct employment or exclusive subcontract — not gig-economy freelancers who fly for a dozen other vendors
Offshore annotation is a cost optimization. Offshore capture is a compliance violation waiting for the right audit.

The Price Differential Is Real And Defensible

U.S.-operated capture crews cost more than offshore alternatives. There is no getting around it. The defense of the price point is that the offshore alternative is not actually available for most of this work — ITAR, site access, clearance, and OPSEC all gate it out before the quote comparison starts. The real comparison is between U.S. crews and "no capture" — and programs that try to substitute "no capture" end up with datasets that don't reflect the operational environment and models that underperform in the field.

The Bottom Line

Defense training data is a U.S. industrial base problem. The crews that capture it need to be part of that industrial base. Vendors that route capture through offshore labor are either doing work that doesn't actually require a cleared crew — in which case the program probably didn't need a premium vendor anyway — or they are creating a compliance problem for everyone downstream. The distinction matters and program offices know it.