This preprint introduces LATTICE, a governance-first architecture for authorized autonomous AI operations. LATTICE enforces deterministic, non-bypassable authorization through a cryptographically verifiable execution gate, separating reasoning from enforcement in autonomous agent systems. The paper formalizes core properties, including traceability, determinism, and invariance, and presents a reference implementation context aligned with high-consequence environments such as cybersecurity testing and critical infrastructure operations.
Elias Calboreanu (Thu,) studied this question.