Lt. Col. Adam Willis, a US Air Force flight surgeon, faces a brutal math problem. An estimated 24% of battlefield deaths in recent conflicts were potentially survivable. Non-compressible torso hemorrhage contributed to 60% of those deaths. The bleeding happens deep inside the chest or abdomen, where no tourniquet can reach, and the clock runs in minutes, not hours. Willis's answer, as program manager for DARPA's new Medics Autonomously Stopping Hemorrhage program, is to train robots to navigate the human torso autonomously, find the bleed, and stop it without a surgeon in the room.
Large-scale battlefield engagements against near-peer adversaries produce casualty counts that overwhelm existing medical evacuation systems. Helicopter extraction to a field hospital, the standard assumption of recent counterinsurgency campaigns, cannot be guaranteed in contested airspace. The calculus is straightforward: if a surgeon cannot reach the casualty within the golden hour, bring the surgeon's capabilities to the casualty inside a robotic platform.
As we wrote in July, the MASH program passed its proposal phase in late 2025, and Phase 1 technical development kicks off this summer. The question is no longer whether autonomous clinical robotics is possible. It is whether the agency can deliver on a problem this hard in 36 months.
Program at a glance
Duration: 36 months (24-month Phase I + 12-month Phase II) · Target: 48-hour pre-hospital stabilization window · Office: DARPA Biological Technologies Office (BTO) · DARPA, 2025
How MASH works: two-phase autonomy architecture
Phase II (12 months) integrates these elements into a fully autonomous workflow. Teams deliver an Objective System, a field-portable prototype designed for Role 1 Military Treatment Facilities, and subject it to independent verification and validation. The system must detect the bleed, navigate to it through a complex field of organs and blood vessels, and deploy a minimally invasive intervention, all without a surgeon directing the sequence.
Our goal with MASH is to give robots guided by advanced sensors, the power of artificial intelligence to locate and stop non-compressible torso hemorrhage, the leading cause of potentially survivable death on the battlefield.— Lt. Col. Adam Willis, DARPA MASH Program Manager
What makes MASH different from every previous medical robotics program at the agency is the autonomy threshold. Earlier systems including the da Vinci surgical robot, which traces its roots to its 1990s investments in medical robotics, kept a human surgeon firmly in control. MASH requires the system to make clinical decisions: where the bleed is, whether to treat it, and which intervention to deploy. The medic is present but functions as an assistant, positioning the robotic platform, changing end effectors, and performing manual scans, while the autonomy software drives the clinical workflow.
This is the convergence point that matters for defence and robotics investors. The agency is not simply funding better surgical tools. It is procuring autonomous clinical decision-making as a formal capability category, sitting alongside weapons systems and targeting algorithms in the Pentagon's R&D portfolio. The contractors best positioned to bid (Boston Dynamics, Shield AI, Sarcos, Medtronic, Stryker) span the defence-robotics and medical-device worlds. The program is designed to pull non-traditional performers into the defence sector: startups, university labs, and first-time defence contractors are explicitly encouraged.
What does MASH unlock for battlefield medicine?
Three structural shifts follow. First, autonomous clinical AI becomes a permanent procurement line at the agency, with adjacent programs feeding into a unified battlefield trauma architecture. Second, the computer vision and robotic navigation problems MASH solves spill directly into industrial robotics and emergency medical services. Third, international humanitarian law faces a novel accountability question: who is liable when an autonomous system makes the wrong clinical decision on the battlefield?
Probability: 60% — the agency has a 36-month track record on similar BTO programs, and the solicitation already attracted multi-disciplinary teams.
The broader trauma portfolio at the agency
Development scenarios
🟢 Optimistic scenario (60%)
Implications: Defence robotics contractors with surgical-robotics divisions (Medtronic, Stryker) gain early positioning. The civilian trauma market opens a new FDA regulatory category for autonomous hemorrhage-control devices.
🟡 Base-case scenario (25%)
Implications: The DARPA-hard framing is accurate but not fatal. Detection-only systems deploy to Role 1 facilities while full autonomous intervention waits for 2032.
🔴 Pessimistic scenario (15%)
Implications: Congress legislates a new liability framework modeled on autonomous weapons rules but with stricter clinical error accountability. Fielding proceeds under restricted rules: autonomous mode only when a medic certifies evacuation is impossible within four hours.