We present Danger-OS, the first application of the Maya affective Spiking Neural Network architecture to a defence and security context. Four Leaky Integrate-and-Fire neurons — Bhaya (fear, τ=3), Vairagya (wisdom, τ=20), Shraddha (trust, τ=10), and Spanda (aliveness, τ=5) — continuously read live system telemetry and produce behavioural anomaly detection decisions at a 500 ms tick cadence. Across eight experimental scenarios totalling 5,710 ticks of continuous operation, we demonstrate: (1) the Bhaya Quiescence Law holds in defence context — aggregate terminal-action rate of 0.315%; (2) reproducibility across independent sessions; (3) adversarial robustness against direct voltage injection; (4) autonomous vigilance from real CPU telemetry; (5) wisdom-mediated SUSPEND vs KILL arbitration; (6) dry-run safety contract — 18 terminal decisions, zero processes terminated; (7) emergent self-monitoring. Paper 1 of the Maya-Defence Series. Extends the 12-paper Maya Research Series (Swaminathan, 2026a–l) into the defence domain. Code private per AI-Defence Working Standards §7.1. Dataset and findings public via this paper.
Venkatesh Swaminathan (Fri,) studied this question.