Containment Before Plasma: Runtime Governance, Unearned Authority, and the Preconditions of Trustworthy Advanced Cognition
Year: 2026
Type: preprint
Abstract: Description Containment Before Plasma presents an architectural approach to a foundational problem in advanced AI governance: how to prevent increasingly capable, probabilistic cognition from silently converting its own fallible reasoning into governing authority. Developed within Architectures of Coherent Intelligence (ACI), the paper argues for a strict separation between proposal and authority, testimony and evidence, candidate state and authoritative activation, implementation and adjudication, and exploration and irreversible consequence. It develops the principles of Zero Unearned Authority, Broad Generation / Narrow Promotion, governed re-entry, evidence-bearing state transition, rollback readiness, and explicit promotion, while proposing a layered view of alignment spanning cognition, epistemics, authority, runtime integrity, persistence, and self-modification. ACI is not presented as a complete solution to AI alignment or superalignment; rather, the paper examines whether trustworthy advanced cognition may require an external governance architecture capable of constraining how cognitive outputs acquire persistence, legitimacy, and real-world consequence.
Author: Petruzzi, Robert
Language: English
Subfield: Safety Research
Field: Social Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7204148800