A defense humanoid mission profile limits a humanoid robot to forward logistics and hazardous-area support: carrying supplies and equipment across rubble and stairs into depots, damaged buildings and contaminated areas, inspecting equipment and scanning buildings with its sensors. It is explicitly never trained for or permitted to use weapons, select targets or make any use-of-force decision. It escalates any contact with people, lost or jammed communications and anything that touches force to a human commander.
The mission profile
| Profile element | This robot |
|---|---|
| Mission | Move supplies and equipment into hazardous areas so people don’t have to go first |
| Operating envelope | Bases, depots, damaged structures, contaminated zones; payload and battery limits enforced |
| Trained for | Load carrying, stair and rubble walking, equipment inspection, sensor scans of buildings |
| Not trained for | Weapons, targeting or any use-of-force decision — excluded by design, not by setting |
| Autonomy level | Acts alone on route, footing and load handling; a commander owns every mission |
| Escalates when | Any contact with people, lost or jammed comms, anything that touches force |
| Learning loop | Every stop, fall and override reviewed before the envelope widens |
What the first field trials actually showed
In March 2026, Foundation sent two Phantom MK-1 humanoids to Ukraine to support frontline forces; the company describes Phantom as the first humanoid built for defense and has received $24 million in research funding from the US Army, Navy and Air Force. Testing results reported in May 2026 were sobering: a payload of about 20 kg, no waterproofing, short battery life, balance problems and malfunctions, vulnerability to jamming and risk if captured. The robots were used mainly for logistics tasks in dangerous areas rather than combat.
That is the case for a mission profile. A general-purpose military humanoid inherits every one of those weaknesses at once. A logistics profile works around them: known loads under the payload limit, short routes inside battery range, a defined response to lost comms — and a hard exclusion on anything involving force.
What it decides, recommends and escalates
| It decides alone | It recommends — a person approves | It stops and escalates |
|---|---|---|
| Route and footing over rubble | Entering a new structure | Any contact with people |
| Load balance and grip | Carrying a load near the payload limit | Lost or jammed communications |
| Returning when battery or comms degrade | Changing the planned route | Anything that touches the use of force |
Where decision intelligence fits
Encodes the red line
The exclusion on force is a policy the platform enforces, logged and auditable — not a setting someone can toggle.
Keeps a named human in command
Every mission has an accountable commander; escalations route to them with sensor evidence attached.
Fails safe
Lost comms or jamming triggers a pre-approved safe behavior — hold, retreat or return — never improvisation.
Widens the envelope only with evidence
New terrain, loads or tasks are added after review of every stop, fall and override.
- Field trials show today’s humanoids suit logistics, not combat.
- The profile excludes weapons and force by design.
- Decision intelligence keeps a named human in command of every mission.