Misalignment Bounty: Crowdsourcing AI Agent Misbehavior

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Turtayev, Rustem, Fedorova, Natalia, Serikov, Oleg, Koldyba, Sergey, Avagyan, Lev, Volkov, Dmitrii
Format: Preprint
Publié: 2025
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866908630834479104
author Turtayev, Rustem
Fedorova, Natalia
Serikov, Oleg
Koldyba, Sergey
Avagyan, Lev
Volkov, Dmitrii
author_facet Turtayev, Rustem
Fedorova, Natalia
Serikov, Oleg
Koldyba, Sergey
Avagyan, Lev
Volkov, Dmitrii
contents Advanced AI systems sometimes act in ways that differ from human intent. To gather clear, reproducible examples, we ran the Misalignment Bounty: a crowdsourced project that collected cases of agents pursuing unintended or unsafe goals. The bounty received 295 submissions, of which nine were awarded. This report explains the program's motivation and evaluation criteria, and walks through the nine winning submissions step by step.
format Preprint
id arxiv_https___arxiv_org_abs_2510_19738
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Misalignment Bounty: Crowdsourcing AI Agent Misbehavior
Turtayev, Rustem
Fedorova, Natalia
Serikov, Oleg
Koldyba, Sergey
Avagyan, Lev
Volkov, Dmitrii
Artificial Intelligence
Advanced AI systems sometimes act in ways that differ from human intent. To gather clear, reproducible examples, we ran the Misalignment Bounty: a crowdsourced project that collected cases of agents pursuing unintended or unsafe goals. The bounty received 295 submissions, of which nine were awarded. This report explains the program's motivation and evaluation criteria, and walks through the nine winning submissions step by step.
title Misalignment Bounty: Crowdsourcing AI Agent Misbehavior
topic Artificial Intelligence
url https://arxiv.org/abs/2510.19738