ALEETH Tests AI Governance System Blocking Unauthorized Actions
ALEETH successfully tested an Institutional Control Architecture that blocked all unauthorized AI agent write and delete attempts during a series of governance trials.
ALEETH announced the successful test of its Institutional Control Architecture (ICA), a governance system designed to prevent unauthorized AI agent actions before they execute. During tests conducted on October 2 and 3, 2026, the ICA blocked all 40 unauthorized write and delete attempts made by NVIDIA Nemotron Nano 12B v2 and Nemotron Super 49B models.
Unlike traditional model-based guardrails, the ICA operates beneath the AI model. It requires signed authority for every action and records all decisions in a public, hash-chained ledger to ensure auditing transparency. The system currently maps controls against 336 frameworks across 53 jurisdictions, including the NIST AI Risk Management Framework and the EU AI Act.
Following these tests, ALEETH launched ALEETH Authority, the first Model Context Protocol server governed by the ICA. The system was developed by The Schreck Group, Inc., the parent company of ALEETH.
