A public evidence registry for qualified robot-policy evaluations, benchmark corrections, and source-verified robot incidents. Ratings and indices remain suspended until a qualified cohort exists.
Physical AI Reliability Index
N/A
unavailable until a qualified learned-policy cohort exists
Benchmark status
—
historical cohort quarantined; no capability comparison is currently valid
Qualified policies
—
Incidents catalogued
—
Qualification controls: —
The Physical AI Reliability Report — chapters, key takeaways, downloadable data.
→◎A transparent record of Physical AI evaluation evidence. The prior D-041 cohort is currently quarantined while Harness v2 is qualified.
→★Ratings are suspended; this page records the qualification gates required before any policy can be rated.
→⚠A source-verified catalogue of real robot harm incidents, cross-linked to primary reports.
→§Public Test & Assessment protocols, operating policies, and the recomputation script.
→