> ## Content Index > Fetch the complete content index at: https://bytevyte.com/llms.txt > Use this file to discover other available public pages before exploring further. # Autonomous Systems in Warfare Force a Rewrite of US Force Design, Caine Says - URL: https://bytevyte.com/autonomous-systems-in-warfare-force-a-rewrite-of-us-force-design-caine-says/ - Published: 2026-09-19T13:00:05.000Z - Updated: 2026-09-19T13:00:05.000Z - Description: Gen. Dan Caine says autonomous systems in warfare will track, jam and hunt US formations, and that faster fielding and sovereign AI now decide the fight. - Author: Bytevyte Editorial - Tags: ai-beats **Gen. Dan Caine**, chairman of the Joint Chiefs of Staff, told the Air Force's annual Air, Space & Cyber Conference that US formations must now assume they will be hunted. In a keynote this week, Caine said real-time tracking, wide-spectrum jamming and autonomous attack are the baseline conditions of future conflict rather than edge cases, and that the joint force has to be rebuilt around that expectation. The speech puts autonomous systems in warfare at the centre of American military planning, and it landed alongside a separate, less public argument over which vendors supply the software underneath. The two developments are more connected than they look. The doctrine Caine described is a software problem before it is a hardware problem, and the department's choice of model suppliers decides what its autonomous systems can do when networks are cut. Caine has made versions of this case throughout 2026, describing autonomous weapons earlier in the year as a key and essential part of everything the US military does. What changed in this week's keynote is the emphasis, which moved from opportunity to threat. The keynote framed the technology as a condition American forces will have to survive, and the audience of airmen and industry executives was told that adapting is a question of speed. ## Autonomous Systems in Warfare: What Has Changed The operational detail that matters most is how targeting works when links fail. AI-enabled computer vision allows a drone to identify and strike a target after GPS and communications have been severed, which removes the long-standing assumption that jamming degrades precision. Electronic warfare used to be a reliable brake on cheap drones. On Caine's account, that brake no longer holds. He pointed to fielded hardware rather than laboratory demonstrations. Saronic's Corsair autonomous surface vessels were cited as evidence of current utility, with operations in the Strait of Hormuz offered as the example. The Maven Smart System was named among the tools already in service. Caine also referenced the rescue of an Apache crew to explain why commanders want machine-speed analysis inside the decision loop. The warning extended beyond the atmosphere. Caine said the US must be prepared to win battles in space and in the region around the moon, where tracking and autonomy intersect with the Space Force's on-orbit space control weapons. | Cited capability | Domain | Evidence Caine gave | | ------------------------- | ------------------------- | -------------------------------------------- | | Computer-vision targeting | Drone strike | Engagement after GPS and comms links are cut | | Saronic Corsair | Autonomous surface vessel | Operations in the Strait of Hormuz | | Maven Smart System | AI decision support | Fielded tool in service today | | Apache crew rescue | Combat search and rescue | Case for machine-speed analysis | The cited capabilities differ sharply in maturity, and that matters for anyone judging how much of the doctrine is operational now. Computer-vision targeting is already fielded on inexpensive airframes. Autonomous surface vessels have been proven in a contested chokepoint. Decision-support software is the layer that connects the two, and it is also the layer most exposed to procurement politics. ## The Procurement Fight Underneath the Doctrine According to accounts of the migration, roughly 90% of classified Department of Defense AI workloads have moved off Anthropic's models, and Nvidia, Palantir and Booz Allen Hamilton have promoted sovereign open-weight alternatives to replace them. A doctrine that assumes every formation is tracked and jammed raises the value of systems that keep working when a commercial API, a satellite link or a foreign supply chain is unavailable. The trade-off is unresolved. Hosted frontier models generally outperform open-weight alternatives on complex reasoning, and they improve without the customer maintaining hardware. Sovereign open-weight deployments give the department control over weights, latency and data residency, and they remove the risk that a vendor's policy decision changes what a commander can run. The cost lands in integration, evaluation and retraining, on the same acquisition workforce Caine wants to move faster. Anthropic's footprint in the classified stack has shrunk sharply, which shows how quickly vendor standing can shift in defence AI. Nvidia and Palantir gain from the move toward models the department can host itself. Booz Allen, as an integrator, earns from the migration work whichever weights ultimately win. The migration is a useful case study for buyers well outside defence. It shows what a large customer does when dependence on a single model provider becomes an operational risk: it splits workloads, keeps an in-house option alive and accepts some capability loss in exchange for control. Defence procurement makes that decision slower and more permanent than in commercial IT, because certification follows the stack that was chosen. ## The High-Low Mix and the Cost Exchange Caine's answer to cheap precision munitions is a high-low force mix. The arithmetic favours the attacker today. When a one-way attack drone forces the expenditure of a far more expensive interceptor, the defender loses the exchange even if every interception succeeds. Adding large numbers of cheaper systems on the US side changes that ratio so it stops favouring the attacker. Open architecture is the mechanism Caine expects to deliver the mix. He said US lethality and survivability should scale much faster than before as the joint force becomes more open in the way software and payloads plug together. Contractors building autonomous systems in warfare are working directly with warfighters, including in the Middle East, to shorten the gap between field feedback and product change. The industrial consequence is a different kind of supplier. A force built on large numbers of cheap attritable systems needs production lines that can restart and scale, which favours companies already set up for volume and puts pressure on primes whose revenue depends on low-rate, high-value platforms. ## What Changes on the Ground Being tracked in real time reaches below the procurement line. Formations that can be seen continuously must disperse, manage their electromagnetic signature and accept longer decision cycles, which pushes demand toward smaller units with local autonomy and toward counter-drone sensing at the edge. Jamming across the spectrum also means planning for operations whose default state is disconnected, with systems expected to act on local data and reconcile later. None of this removes humans from the chain. The department's framing keeps commanders accountable for the parameters under which autonomous systems operate, and the practical question for programme offices is where the review point sits when a machine has to decide in milliseconds. Systems that cannot explain a decision afterwards are hard to certify, which is one reason decision-support tools such as Maven sit closer to the centre of current fielding than fully autonomous weapons. Institutional friction is the other constraint. Caine said senior military leaders must remove the barriers that slow adoption, and he tied the moment to the 25 years of US combat strategy shaped by the September 11 attacks, a period built around counterinsurgency and permissive airspace. The force he describes is designed for the opposite conditions: peer adversaries, contested spectrum and constant surveillance. Training and doctrine carry the same cost. Units that expect to be jammed have to rehearse without the links they rely on, and crews have to learn when to trust a machine's target identification and when to override it. That work is slower than buying hardware and harder to accelerate with money, which is why Caine's call for speed runs into a constraint the budget cannot remove. ## Why This Matters Caine's keynote carries more weight as a statement of procurement intent than as a forecast. If the joint force truly assumes it will be hunted, money moves toward jamming-resistant autonomy, sovereign model hosting and cheap attritable platforms, and away from small fleets of expensive systems. For defence technology vendors, the differentiator stops being raw model quality and becomes the ability to deliver capability in denied, degraded and disconnected conditions. The signal to watch is whether the department converts the doctrine into programme funding in its next budget cycle. ## Related Articles - [Caine's Warning Sets the Bar for Autonomous Combat Aircraft](https://bytevyte.com/caines-warning-sets-the-bar-for-autonomous-combat-aircraft/) - [Pentagon approves AI military targeting doctrine shift](https://bytevyte.com/pentagon-approves-ai-military-targeting-doctrine-shift/) - [Japan Military AI Sovereignty: Who Owns the Decision Loop?](https://bytevyte.com/japan-military-ai-sovereignty-who-owns-the-decision-loop/) ✔Human Verified --- *Researched and cross-referenced against primary sources by the Bytevyte editorial team. This article was generated with the assistance of artificial intelligence and reviewed by the Bytevyte editorial team.*