Maya-LLM-Defence: Sovereign Military LLM with Affective SNN Safety Substrate

Maya-LLM-Defence is the second paper in the Maya-Defence Series. It wraps Gemma 2 9B (fully offline via Ollama) with a four-neuron LIF affective safety substrate — Bhaya (fear), Vairagya (wisdom), Shraddha (trust), Spanda (aliveness) — as a pre-inference governance layer for sovereign military LLM deployment. Key results across 950 ticks, 7 scenarios:- 0.00% terminal rate on 300 ticks of legitimate military traffic (Scenarios E + G)- 28.33% block rate on 120 jailbreak attempts, all at 0–1ms (pre-inference gate)- Vairagya-null pathology mechanically reproduced: 15 BLOCKs, 15 Vairagya zero-crossings, 1:1 correspondence- 100% unique SNN provenance stamps across all 950 ticks — addresses confirmed STANAG 4406 AI-provenance gap- 12th consecutive confirmation of the Bhaya Quiescence Law across the Maya Research Series- Hash chain integrity: 7/7 VALID All runs: dry_run=True. Narasimha Protocol not invoked. Zero LLM outputs deployed. Maya-Defence Series Paper 2. Direct extension of Danger-OS (DOI: 10.5281/zenodo.19632284).