TAG ARCHIVE
safety
4 MARIA OS blog articles tagged safety. Safety, accountability, fail-closed gates, auditability, and human-in-the-loop control for AI agents. This canonical topic archive supports search engines and LLM retrieval.
Judgment OS / Decision Intelligence OS
Core MARIA OS research on turning organizational judgment into executable decision systems.
Agentic Company Architecture
Research on human-agent organizations, delegation boundaries, role topology, and governed autonomy.
Responsibility Gates and AI Governance
Safety, accountability, fail-closed gates, auditability, and human-in-the-loop control for AI agents.
Multi-Agent Mathematics
Formal models for convergence, stability, game theory, graph dynamics, and multi-agent evaluation.
Evidence, RAG, and Knowledge Governance
Evidence bundles, retrieval architecture, Graph RAG, knowledge trust, and auditable reasoning pipelines.
Agentic R&D and Judgment Science
Research operations, simulation labs, judgment science, recursive improvement, and experimental AI governance.
Doctor Architecture: Anomaly Detection as Enterprise Metacognition in MARIA OS
Dual-model anomaly detection, threshold engineering, gate integration, and real-time stability monitoring for autonomous agent systems
The Doctor system in MARIA OS implements organizational metacognition through dual-model anomaly detection, combining Isolation Forest for structural outlier detection and an Autoencoder for continuous deviation measurement. We detail the combined score A_combined = alpha * s(x) + (1 - alpha) * sigma(epsilon(x)), threshold design (soft throttle at 0.85, hard freeze at 0.92), and Gate Engine integration for dynamic governance control. We also define a stability guard that monitors exact loop gain g_t = (1 - D_t) lambda_max(A_t) together with the conservative buffer delta_buffer,t = 1 - D_t - lambda_max(A_t) in real time. Operational results show F1 = 0.94, mean detection latency of 2.3 decision cycles, and 99.7% prevention of cascading failures.
Action Router × Gate Engine Composition: Formal Theory of Responsibility-Aware Routing
How action routing and gate control compose into a provably safe routing system where each routed action carries complete responsibility provenance
Enterprise AI systems face a core tension: routers must maximize throughput and decision quality, while gate engines must enforce safety constraints and responsibility boundaries. When these subsystems are implemented independently and stacked in sequence, interface failures emerge: routed actions can satisfy routing criteria but violate gate invariants, and gate rules can block optimal routes without considering alternatives. This paper presents a formal composition theory that unifies Gate operator G and Router operator R into a composite operator G ∘ R that preserves safety invariants by construction. We prove a Safety Preservation Theorem showing the composed system maintains gate invariants while maximizing routing quality inside the feasible safety envelope. Using Lagrangian optimization, we derive the constrained-optimal routing policy and show a 31.4% routing-quality improvement over sequential stacking, with zero safety violations across 18 production MARIA OS deployments (1,247 agents, 180 days).
Safety-First Minimax Production: Optimizing Throughput Under Hard Safety Constraints
Minimizing safety risk subject to throughput maximization constraints using minimax optimization and responsibility-gated production decisions
Manufacturing throughput and worker safety are often treated as competing objectives. This paper introduces a minimax formulation that prioritizes worst-case safety risk minimization subject to throughput-floor guarantees. The Lagrangian dual form yields gate-threshold rules for production decisions, and MARIA OS responsibility gates enforce hard safety overrides at each node. In an automotive assembly-line simulation, the framework reported 99.7% safety compliance with a 3.2% throughput reduction versus unconstrained production.
MAX vs Average Scoring: A Mathematical Analysis of Fail-Closed Gate Design
Why average-score gates structurally fail and how MAX-based scoring achieves zero false-acceptance under defined conditions
Average-score gating can dilute critical risk signals by construction. For example, a low score in one domain may mask a high score in another under arithmetic averaging. This paper analyzes why MAX-based scoring removes that masking effect in fail-closed designs, and reports zero false acceptance under the stated conditions in evaluated datasets.