tiers

Research

Papers from tiers on the governance and economics of AI inference.

Preprint, September 12, 2026

Governed Before Execution: The Hugging Face Intrusion as a Case Study in Trajectory Governance, with Evidence from 2024 to September 2026

The July 2026 intrusion passed through four simpler violations of its declared contract before the first remote code execution. This paper maps the public record, and two further incidents disclosed the same week, against a four-function governance architecture and classifies what a governor on the execution path would have prevented, contained, detected earlier, or could not have seen.

AbstractPDFBibTeXRIS

Research paper, September 8, 2026

The anatomy of an agent dollar

An accounting framework for autonomous work: seven destinations for agent spend, three simultaneous ledgers, and metrics that connect trajectory cost to verified outcomes.

AbstractMarkdownBibTeXRIS

White paper, August 25, 2026

The Intelligence Record: An Empirical Case for AI Inference Governance

The architecture: four functions acting on every call, the session as the security-relevant runtime, verification against declared intent, and a signed receipt closing every governed disposition. With the signed counterfactual methodology and the empirical case behind it.

AbstractPDFBibTeXRIS