Saved in:
Bibliographic Details
Main Author: Steger, Cynthia
Format: Recurso digital
Language:English
Published: Zenodo 2025
Subjects:
Online Access:https://doi.org/10.5281/zenodo.18110896
Tags: Add Tag
No Tags, Be the first to tag this record!
Table of Contents:
  • <p>Current AI safety approaches treat symptoms (hallucinations, misuse, mental health impact) as behavior problems requiring external restrictions. This brief proposes they are architecture description problems requiring accurate system self-reference and user understanding.<br>Three interconnected safety challenges—hallucinations, user psychosis, and infrastructure harm—share one root cause and one solution: inaccurate description of what AI systems are.</p>