Openwist

Nia by Openwist is a supervisory safety layer for machines that decide for themselves: it sits between a learned policy and the robot, checks every proposed action against explicit human-approved rules, and under tighten-only authority can pass the action, substitute a safe fallback, or hand over — never anything else. Every decision becomes signed, tamper-evident evidence. On pre-registered held-out runs: standby stops in 1.8 s (vs 8.2 s), sensor-dropout contact 84% to 0, untrusted-command falls 100% to 0, policy-fault falls to 0% after a disclosed fix, nominal operation untouched. A Muse advisor (Muse Spark 1.3) drafts cited incident reports and rule proposals that must clear regression and human signing. Simulation only; Meta Motivo is the designated third-party test subject. 189 tests green; pip install nia-safety.