Thousands of U.S. and allied troops recently converged on the California desert to stress-test artificial intelligence-driven command and control systems in one of the largest joint military experiments of its kind. The exercise, known as Project Convergence Capstone 5, brought together personnel from multiple partner nations at Fort Irwin, California, to evaluate whether machine-assisted decision tools can accelerate the path from detecting a threat to engaging it. The results could reshape how allied forces coordinate across air, land, sea, space, and cyber domains in future conflicts.
Why AI-driven command experiments at Fort Irwin demand attention now
The core question driving these exercises is speed. Modern warfare increasingly rewards the side that can process sensor data, identify targets, and authorize strikes faster than its opponent. Project Convergence exists to test whether AI-enabled links between allied forces can compress that timeline enough to matter against a technologically advanced adversary such as China or Russia.
Capstone 5 at Fort Irwin represents the latest iteration of what the Pentagon has structured as persistent experimentation tied to joint all-domain command and control concepts. Each successive Capstone event builds on the previous one, adding new technologies, more participants, and harder scenarios. The progression from Capstone 4 to Capstone 5 is designed to push integration further, testing whether systems that worked in controlled settings can function under realistic battlefield pressure with live troops maneuvering across harsh terrain.
A working hypothesis for these exercises is that multinational AI command links could reduce the sensor-to-shooter timeline by a significant margin compared with earlier Capstone iterations. No official after-action report has yet confirmed a specific percentage improvement. Verifying that claim will require access to future performance metrics from the Department of Defense, which typically releases findings months after each event concludes. The absence of published data from Capstone 5 means that any quantified improvement remains unconfirmed.
What is already clear is that the stakes extend well beyond a training range in the Mojave Desert. If these AI tools prove reliable in connecting allied forces across domains and national boundaries, they could change the calculus for how the United States and its partners plan to fight together. If they fail or introduce new vulnerabilities, the investment of years of development and billions in defense spending faces hard questions.
Allied forces and AI tools tested at Capstone 5
Australia’s participation at Capstone 5 offers one of the clearest windows into how allied nations are folding their own capabilities into the U.S.-led experiment. The Australian defence ministry published reporting on its role at Fort Irwin, describing the integration of emerging technologies into live desert training alongside American forces. Australian officials framed the effort as focused on converting analytical capability into actionable firepower, a goal that depends on AI systems sorting through massive volumes of sensor data and presenting options to human commanders fast enough to act on them.
The exercise brought together forces from multiple allied nations, though exact troop counts and unit designations for every participating country have not been disclosed in the primary institutional releases. Australia provided concrete personnel figures in its own reporting, but aggregate numbers across all nations remain unspecified in publicly available documents. That lack of granularity makes it difficult to quantify the scale of multinational participation, even as officials emphasize the breadth of the coalition involved.
Project Convergence’s design reflects a deliberate choice by the U.S. Army and the broader Defense Department to treat these exercises not as one-off demonstrations but as a continuous testing pipeline. Each Capstone event layers new challenges onto the previous baseline. Capstone 4, for example, explored how joint forces could share targeting data across services and domains. Capstone 5 expanded that ambition to include deeper allied integration and more advanced AI decision-support tools operating under field conditions.
The practical mechanics matter here for anyone watching defense procurement or alliance politics. When an Australian soldier’s sensor feed connects through an AI system to an American fires platform in real time, the technical plumbing behind that exchange involves interoperability standards, data-sharing agreements, and trust frameworks that take years to negotiate. Success at Fort Irwin does not automatically translate to success in a contested environment where communications are jammed and satellites are targeted, but it establishes a baseline that future planning depends on.
For allied militaries, Capstone 5 also served as a rehearsal for the bureaucratic and legal work required to make AI-enabled cooperation possible. Information assurance rules, classification constraints, and national caveats can all slow or block the sharing of data that AI tools rely on. Testing those boundaries in an exercise environment helps identify where policies must change if commanders are to benefit from machine-speed analysis in real operations.
Risks, limits, and human control
Underlying the technical experiments is a debate about how much authority AI systems should wield in wartime. Officials involved in Project Convergence have consistently stressed that humans remain in the loop for lethal decisions, with algorithms providing recommendations rather than orders. Capstone 5 continued that pattern by focusing on decision support, not autonomous engagement.
Nonetheless, compressing decision timelines carries inherent risks. If AI tools misclassify a target or fail to account for civilian presence, the speed they enable could magnify the consequences of error. Exercises like Capstone 5 are partly intended to surface those failure modes in a controlled setting, allowing engineers and commanders to refine safeguards before any system is fielded at scale.
There is also the question of resilience. The AI-enabled networks tested at Fort Irwin depend on reliable connectivity, robust data flows, and secure software. Adversaries will seek to disrupt or deceive those systems through jamming, cyber attacks, and spoofed signals. Capstone 5 provided an opportunity to probe how AI tools perform when inputs are degraded or corrupted, but the public record does not yet detail how aggressively those stressors were applied or how the systems responded.
Gaps in the Capstone 5 record and what to watch next
Several significant questions remain unanswered in the public record. The most consequential is performance data. Neither the U.S. Department of Defense nor allied governments have released specific metrics showing how much faster AI-assisted command loops operated compared with traditional methods during Capstone 5. Without those numbers, the central promise of the exercise, that machine speed can give allied forces a decisive edge, rests on institutional assertions rather than demonstrated results.
The names and capabilities of specific AI systems used during the exercise are also absent from the institutional releases reviewed. Defense officials have historically been cautious about identifying individual software platforms tested during Project Convergence, partly for security reasons and partly because many tools are still in development. That opacity makes independent assessment difficult and complicates efforts by outside analysts to track which technologies are advancing toward operational use.
Direct statements from senior U.S. or allied commanders about observed decision-time improvements during Capstone 5 have not appeared in the primary materials. Previous Capstone events produced some public commentary from generals and program managers, but detailed after-action findings typically surface well after the exercises end, sometimes taking months to clear review processes. Until those assessments are released, outside observers must rely on broad characterizations rather than quantified outcomes.
In the coming months, several indicators will show whether Capstone 5 meaningfully advanced the state of allied AI-enabled warfare. Budget documents and acquisition plans may reveal which experimental tools are being fast-tracked for further development. Policy updates on data sharing and interoperability could signal that lessons from Fort Irwin are reshaping how allies plan to fight together. Most importantly, any eventual release of performance metrics will help clarify whether the promise of faster, more integrated decision-making is being realized or remains aspirational.
For now, Capstone 5 stands as a high-profile waypoint in a longer campaign to adapt alliance command structures to an era of ubiquitous sensors and machine intelligence. The exercise underscored both the potential of AI to transform coalition operations and the many unanswered questions about reliability, accountability, and real-world effectiveness. How governments address those questions will determine whether the experiments in the Mojave Desert become the foundation of future warfighting or a cautionary tale about the limits of battlefield automation.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
