Who holds the reins when artificial intelligence moves from tool to teammate on the battlefield — the uniformed commander or the statistical model itself? That dilemma lies at the heart of a recent examination by former Department of Defense chief information officer Leslie Beavers, who frames ethics, reliability, vendor risk and autonomy as central questions as AI becomes a core national security instrument.
Background: AI moving into defense operations
Leslie Beavers, who served as the Department of Defense chief information officer, assesses a rapidly shifting operating picture: AI is increasingly used in defense operations, and the technology is evolving from an adjunct system into what Beavers describes as a key combatant. That shift forces a simple but uncomfortable question — who controls the system: the military or the model?
The control dilemma: ethics, reliability and autonomy
Beavers highlights three interlocking challenges. First, ethical considerations arise as decision-making that once rested squarely with people is ceded, in part or whole, to algorithmic systems. Second, reliability becomes paramount when AI outputs can alter the course of operations. Third, autonomy — the degree to which systems act without human intervention — changes the nature of command, oversight and responsibility.
Vendor risk and competing perspectives
Beavers also explores vendor risk as another axis of concern. As military organizations rely on external developers and commercial products, questions surface about influence, supply-chain integrity and alignment between operational needs and model behavior. In her analysis, these risks intersect differently depending on the observer:
- Technologists: Focus on model behavior, failure modes and engineering controls that can make AI outputs more predictable and auditable.
- Policymakers: Confront questions of governance, accountability and the legal and ethical frameworks needed to govern algorithmic action in conflict.
- Users — military personnel: Face the practical tradeoffs between speed, trust and human-in-the-loop control when systems are deployed in high-stakes environments.
- Adversaries: Present a factor that reshapes incentives — the prospect that opponents may exploit model weaknesses or mirror AI-enabled tactics.
Why it matters
Beavers’ account frames the issue as more than a technical or procurement problem: it is a national security question about control, predictability and values. If AI systems are to be treated as combatants in their own right, the mechanisms for oversight, fail-safe controls and vendor accountability will determine whether those systems augment human judgment or supplant it in ways that create new risks.
The path forward, in Beavers’ assessment, requires confronting these tradeoffs explicitly: designing for reliability and ethics, managing vendor relationships responsibly, and clarifying how autonomy will be bounded in operational contexts. Absent that disciplined approach, the answer to “who controls the system” risks being decided by the contours of the models themselves rather than by the people and institutions charged with national defense.
As AI presses deeper into military operations, the enduring question remains: will policy, oversight and engineering keep control in human hands, or will the model — by design or default — write the rules of engagement?
https://www.govinfosecurity.com/who-controls-ai-on-battlefields-military-or-model-a-31379