Illustrative telemetry from a demonstration run: localization ok at ±2 cm; obstacle tracked at 1.8 m; grasp planned at conf 0.96; then an unfamiliar scene at conf 0.41 triggers operator takeover.
Physical AI · early access
Insurance-grade telemetry
for Physical AI.
Safety envelope, hardware vitals, autonomy — metered per run and graded per fleet, so humanoids, AMRs, and arms can be insured on measured risk.
The three signals that decide whether a robot is insurable.
How close, how fast, how hard.
Human proximity, speed, and contact force — logged every run, as evidence.
The machine's own health.
Torque, temperature, battery — mechanical failure signals early, so we meter it early.
How often humans step in.
Every operator takeover counted — the cleanest maturity signal a fleet produces.
Measured in the language of ISO 10218 and ISO/TS 15066.
Metered per run. Graded per fleet. Priced on evidence.
Metering on your stack
A drop-in telemetry layer for your robot software stack — the physical-AI counterpart of the AgentRisk battery.
In designFleet exposure grade
Every run scored, every unit graded, trends across the fleet — the same content-free scorecard discipline as our agent ratings.
PrototypeCoverage on measured risk
Bodily-injury and property-damage exposure priced from what the fleet actually does — not a static questionnaire.
Building with partnersBuilt in the open, with early partners.
The AgentRisk rating for AI agents is live today at app.tryauly.com. Physical-AI metering is in design with early partners — robot OEMs, fleet operators, and insurers shaping what gets measured first.
Or write to us: support@kalta.ai — tell us what you run and where it operates.