Saved in:
| Main Author: | |
|---|---|
| Format: | Recurso digital |
| Language: | |
| Published: |
Zenodo
2026
|
| Online Access: | https://doi.org/10.5281/zenodo.19570179 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Table of Contents:
- <p>Agenity Research Series — Paper 6A. This paper introduces the Agent Identity Gate (AIG), the enforcement layer for the identity infrastructure established in Paper 1. Motivated by Stein et al. (2026, arXiv:2604.11806), who demonstrate that anonymous agent operation enables distributed safety violations detectable only through expensive post-hoc forensic auditing, we propose that unregistered agents be restricted to local model execution while frontier AI API access requires a verified, cryptographically signed Agent Identity Token (AIT). The paper specifies the full AIG architecture: federated Registration Authority, AIT token schema, scaffold attestation, gateway enforcement, and ledger binding. Three open problems from Meerkat — adversarial embedding attacks, model-agnostic scaffold cheating, and cross-trace hyperproperty violations — are structurally addressed. Paper 6A is the direct prerequisite for Paper 6B (Agent Labor Markets): an agent without verified identity cannot participate in the Agenity economy.</p>