Opinion & Analysis
Opinion

Biodata, AI and the future of command: Getting it right

Fitness trackers and AI diagnostics are here to stay. They can provide valuable data for operational planning — if security and privacy are handled properly.

Sailors attached to Expeditionary Medical Force One participate in an emergency room simulation during Operation Northern Lights. Northern Lights is an annual Navy Reserve Medical training exercise hosted at Fort McCoy, Wis. by Expeditionary Medical Facility Great Lakes One, a Navy Reserve expeditionary medical unit based out of Great Lakes, Ill. The purpose of this training exercise is to provide practical experience in the Expeditionary Medical Facility environment. There is classroom instruction, simulated training and practical experience in field hospital operation and patient care services. (US Navy photo by Mass Communication Specialist 2nd Class John Sorensen/Released).

As AI and biotech converge, the central question is no longer simply whether the military can collect biological data. The question is how.

Who controls collection? Under what conditions may the data be transmitted? Who is permitted to interpret them? What authority should those interpreters carry in command decisions? I contend that the Pentagon — and, in the last resort, Congress — must govern biodata as an emerging factor in not just military medicine, but in command decisions. Its confidentiality, integrity, availability and operational use are increasingly matters of force protection and command responsibility.

Emerging technologies can collect and interpret indicators of fatigue, cognitive overload, physiological stress, environmental exposure and recovery. The underlying data may include heart rate and variability, sleep and recovery measures, body temperature, blood chemistry, biomarkers of injury or infection and indicators of physiological stress, as well as less obviously medical metrics like movement and activity patterns. 

Used responsibly, these capabilities could help commanders anticipate declining endurance, manage force rotations, prioritize limited medical resources and make better-informed decisions about operational readiness. But the same information, in the wrong hands or just mishandled, can expose a unit’s vulnerabilities or distort the decisions on which its mission depends.

There is a critical distinction to make here between raw data and what is done with it, by whom. A sensor measures something; an AI system interprets what that measurement may mean; a commander decides whether that assessment matters to a particular mission. These are different functions, carrying different standards of reliability and responsibility. A heart-rate measurement is not a diagnosis. A sleep deficit is not a complete measure of judgment. A model-generated readiness score is not an order, nor should it become one by default. Human judgment remains essential and humans must be held accountable for their judgments — or for their blind deference to a machine.

So how can the military use biodata right? To start with, during training, biodata could help identify accumulated fatigue, optimize recovery schedules, evaluate exposure to environmental hazards and support earlier medical assessment. Longitudinal data collected over months and years can establish an individual’s baseline, allowing clinicians to assess deviations from that person’s individual range, rather than compare every service member against an undifferentiated population average. Such uses can improve health, training design and force preservation before a deployment begins.

Moving beyond individual health, AI could identify changes in performance across a unit, help anticipate demand for medical services, or inform decisions about when a unit needs to rest and when it can keep operating. When fused with other, more traditional military data — like personnel availability, ammunition stocks, equipment status, intelligence and mission requirements — biodata becomes part of the commander’s broader operational picture.

That increases both the utility and the risk. Medical information becomes command information when an inference about the body begins to influence where, when and how a person or formation is employed. At that point, the military must be able to distinguish legitimate operational awareness from inappropriate surveillance, and decision support from human judgment blindly bowing to a statistical estimate.

Security And Privacy

Before a commander asks what a wearable device can reveal about personnel, a more basic question must be answered: Should that device be active in this environment at all?

In 2018, the Defense Department restricted geolocation capabilities in designated operational areas, recognizing that fitness trackers and related devices could expose personnel locations, routines and force concentrations. 

Even data collected during peacetime can create a security risk. Over time, measurements may reveal who trains together, how long units take to recover, how long personnel remain at a facility and when readiness appears to deteriorate. A persistent record of routine behavior can become a map of future military capability — for our own planning or, if leaked, for an adversary’s.

Precisely because biodata is operationally useful, it is also worth stealing or manipulating. An adversary able to access physiological, behavioral and performance data could identify training cycles, recovery windows, deployment patterns, or possible gaps in unit endurance. Or, instead of just stealing biodata, an adversary could alter it, making a capable formation appear unfit for operations or portraying an exhausted unit as ready, which could gaslight commanders to push tired personnel beyond their limits or prematurely withdraw high-functioning forces.

Even security measures can be revealing: If everyone in a given unit turns off their trackers at once, that itself is a signal they’re likely about to be deployed. Restrictions on devices must cover the entire life cycle of the data, not simply the moment of contact.

In short, a personal health tracker may have implications well beyond its individual user — and simply turning off location sharing, or setting a device on “airplane mode,” does not necessarily eliminate the risk.

So the question for commanders is not simply whether a device is approved for purchase or useful during training. It is who has authority to permit, forbid, or restrict the device when security conditions change.  A tracker that is valuable during a rehabilitation program or garrison training may be unacceptable near a sensitive installation, during preparations for deployment, or in a conflict zone. Permissions should be specific to the mission and environment, with explicit distinctions between collecting data locally, transmitting it wirelessly, and uploading it to cloud services. In some settings, the correct command decision will be to leave the device behind.

When Medical Information Becomes Command Authority

A second set of questions concerns permissions, identifiability and interpretation: Who can do what, when? Can medical professionals, military superiors and the service members themselves all see the same information? May a commander access an individual’s clinical indicators, or only an operationally relevant summary? Who may share, retain, challenge, or delete a record?

These distinctions must be designed into the policy and the technology, not tacked on ad hoc. Medical staff may need a servicemember’s personal, identifiable information for diagnosis and treatment. A commander may only need a bounded assessment of mission-relevant limitations like fatigue or illness, without access to a full medical history. The individual should have a way to understand how their information is used —and to contest misuse. 

Baseline comparisons create another challenge. Comparing a service member with their own validated baseline may be more informative than measuring them against generalized averages across large populations. But illness, local environments, new equipment and incomplete measurements can alter what counts as “normal.” Rather than rubberstamp assessments they don’t understand, commanders should know whether any given assessment reflects a direct measurement, a deviation from an individual baseline, a population-based inference, or an AI-generated prediction and whether the model has been validated for the mission at hand.

AI can illuminate aspects of the problem, but it cannot by itself settle questions of unit cohesion, tactical adaptability, or military necessity. Users must guard against “automation bias,” where humans defer uncritically to computer outputs as if to an oracle, and “metric fixation,” where the ability to measure a few factors leads to ignoring important intangibles. Biological measurements should inform, not silently substitute for, professional military judgment and clinical expertise.

An AI-generated finding can appear more precise than the underlying evidence actually is. Consider a dashboard reporting that a service member is “90 percent ready.” Ready for what task, for how long, under what environmental conditions and according to which validated criteria? A numerical score may compress complex uncertainties into an illusion of simplistic certainty.

Assessments should therefore distinguish between actual measurements, diagnoses based on that hard data and predictions based on those diagnoses, with clear caveats about the uncertainty and limits inherent in any mathematical model. They should also permit a human to challenge and correct potential errors. A clinical judgment, a service member’s report, or a commander’s direct observation may reveal conditions that the model did not capture.

Recommendations: Securing the Biodata Architecture

Overall, a coherent approach to biodata should treat the following factors, not as afterthoughts, but as central to the design of the system:

First, data collection and emissions control must be explicit command decisions. Establish mission-specific rules, designate the authority to issue exceptions and require documented risk assessments for sensitive operations or locations. Build the system to degrade gracefully instead of collapsing as soon as an enemy disrupts the network. Never assume that equipment suitable for peacetime training is suitable for operational deployment.

Second, protect biodata for both individual privacy and operational security. Apply safeguards proportionate to the mission context and the value an adversary could derive from aggregation and correlation with other data. Secure the data pathway end-to-end, starting with vetted sensors and ending with strict rules for long-term retention or storage.

Third, separate clinical access from command use. Establish clear and different levels of access for medical personnel, commanders, analysts and service members. Military leaders should receive the minimum personal information necessary for legitimate decisions, while clinical determinations remain with medical professionals. Specify who may activate monitoring systems, inspect identifiable records, approve secondary uses and audit access.

Fourth, guard against blind trust in algorithmic claims. Require outputs to distinguish individual baselines from population norms, direct observations from predictions and validated results from uncertain inferences. Build in procedures for challenging a score, documenting overrides, correcting errors and preventing unjustified decisions based on incomplete data.

Fifth, assign accountability before operational adoption. Identify who approves new technologies, sets performance thresholds, adjudicates conflicting assessments, reviews risks and investigates failures. Responsibility must remain traceable to identifiable, individual humans.

Finally, train for biodata denial, deception and manipulation. In wargames, present commanders with seemingly credible readiness estimates that conflict with what their medical team, their subordinates, or their own eyes are telling them. Scenarios should include personal trackers transmitting during an exercise, improperly anonymized records exposing a small unit, or an AI tool producing corrupted conclusions. The objective is to develop judgment about when to trust an assessment, when to question it and when to operate without it.

The United States has an opportunity to use AI and biotechnology to strengthen medical care, personnel readiness and military effectiveness. Realizing that potential requires more than deploying sensors or buying analytics. It requires institutions capable of governing when information is created, where it flows, who can interpret it and how much authority it should have over consequential decisions.

Protecting both individual warfighters and the force as a whole requires protecting both the information drawn from the human body and human authority over how it is used. Biodata security is not simply a privacy issue or a medical-technology problem. It is a requirement of force protection, operational resilience and leadership.

Dr. Elise G. Annett is a Research Fellow in the Program for Disruptive Technology and Future Warfare at the Institute for National Strategic Studies, National Defense University, and a Senior Digital Domain Fellow at the Center for the Application of Grand Strategy.

The views and opinions presented in this essay are those of the author and do not necessarily represent those of the United States government, the Department of War, or the National Defense University.