A Mereotopological Cellular Automata Architecture and Method to Demarcate LLM Operational Boundaries

This white paper proposes the Modular Cellular Automaton Model (MCAM) as a framework for enforcing safety boundaries on AI systems. Instead of attempting to align AI internally with human values, MCAM constrains behavior externally by evaluating proposed actions against spatial and semantic rules encoded in a cellular-automaton grid. By labeling locations (such as critical infrastructure) and applying rule-based checks before execution, the system functions as a semantic firewall that blocks unsafe actions while providing transparent, auditable enforcement of operational limits. Download it here.