Emergency Intervention and Verification in AI-Dependent Infrastructure

Stopping an AI deployment does not necessarily revoke the authority it has acquired through other systems. This provisional preprint proposes a bounded assessment procedure for evaluating whether two emergency intervention paths remain effective under specified failures or compromises, distinguishing shared components from shared defeat conditions and separating successful action from independently verifiable evidence of the protective outcome. It yields a scoped pass, failure or inconclusive result rather than a claim of statistical independence or universal controllability. A short assessment of electromagnetic-pulse proposals and a U.S.-only legal survey explain the conditions under which the test can be used. Status: this is a provisional proposed method. It has not been peer reviewed, has not been tested by independent assessors, and has not been validated in real facilities. The worked example (a greenhouse-control laboratory, with all timings, observations and assessment records) is entirely fictional and illustrative; it is not evidence from an actual facility. The paper builds on the intervention requirement discussed in Li (2026), Unkillable (https://doi.org/10.5281/zenodo.23223075), and on the observation-window concept in Li (2026), Which Failure Do You Fear More? (https://doi.org/10.5281/zenodo.22821194). AI assistants helped with drafting, editing, citation checking and formatting; the author directed and reviewed all content and is responsible for it.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-17
DOI
https://doi.org/10.5281/zenodo.24323862
Primary Topic
Safety Systems Engineering in Autonomy
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

Emergency Intervention and Verification in AI-Dependent Infrastructure

Ana (李安娜) Li
4 links
Zenodo (CERN European Organization for Nuclear Research)
Safety Systems Engineering in Autonomy
preprint

Emergency Intervention and Verification in AI-Dependent Infrastructure

Ana (李安娜) Li
preprint en

Abstract

Stopping an AI deployment does not necessarily revoke the authority it has acquired through other systems. This provisional preprint proposes a bounded assessment procedure for evaluating whether two emergency intervention paths remain effective under specified failures or compromises, distinguishing shared components from shared defeat conditions and separating successful action from independently verifiable evidence of the protective outcome. It yields a scoped pass, failure or inconclusive result rather than a claim of statistical independence or universal controllability. A short assessment of electromagnetic-pulse proposals and a U.S.-only legal survey explain the conditions under which the test can be used. Status: this is a provisional proposed method. It has not been peer reviewed, has not been tested by independent assessors, and has not been validated in real facilities. The worked example (a greenhouse-control laboratory, with all timings, observations and assessment records) is entirely fictional and illustrative; it is not evidence from an actual facility. The paper builds on the intervention requirement discussed in Li (2026), Unkillable (https://doi.org/10.5281/zenodo.23223075), and on the observation-window concept in Li (2026), Which Failure Do You Fear More? (https://doi.org/10.5281/zenodo.22821194). AI assistants helped with drafting, editing, citation checking and formatting; the author directed and reviewed all content and is responsible for it.

Zenodo (CERN European Organization for Nuclear Research)
Peace, justice, and strong institutions
Safety Systems Engineering in Autonomy
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon