Saved in:
Bibliographic Details
Main Authors: Goemans, Arthur, Altman, Dan, Dreksler, Noemi, Freund, Jonas, Gandhi, Milan, Wang, Zhengdong, Cogan, Sarah, Krier, Sebastien, Brady, Demetra, Ho, Lewis, Dafoe, Allan
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2606.00047
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916067816767488
author Goemans, Arthur
Altman, Dan
Dreksler, Noemi
Freund, Jonas
Gandhi, Milan
Wang, Zhengdong
Cogan, Sarah
Krier, Sebastien
Brady, Demetra
Ho, Lewis
Dafoe, Allan
author_facet Goemans, Arthur
Altman, Dan
Dreksler, Noemi
Freund, Jonas
Gandhi, Milan
Wang, Zhengdong
Cogan, Sarah
Krier, Sebastien
Brady, Demetra
Ho, Lewis
Dafoe, Allan
contents Frontier AI governance often centres on the model-level governance paradigm, which assumes that a model's capability profile is primarily a function of the compute and data used during training. This position paper argues that model-level governance becomes less effective when capability progress is increasingly driven by "non-model gains"--improvements that are independent from advances in the base model. We formalise the concept of non-model gains and provide a taxonomy of three distinct vectors of capability gain: inference gain (scaling compute at test-time), systems gain (post-training enhancements such as scaffolds), and asset gain (enhancing a model with restricted assets). We demonstrate how these vectors--alongside potential future impacts from embodiment, continual learning, and AI diffusion--may undermine risk management strategies that hinge mostly on pre-deployment evaluation and mitigation. We provide an overview of governance approaches that go beyond the model level: system, entity, agent, and cloud governance. Finally, we emphasise the importance of societal resilience as a complement to these governance layers.
format Preprint
id arxiv_https___arxiv_org_abs_2606_00047
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Comprehensive AI governance requires addressing non-model gains
Goemans, Arthur
Altman, Dan
Dreksler, Noemi
Freund, Jonas
Gandhi, Milan
Wang, Zhengdong
Cogan, Sarah
Krier, Sebastien
Brady, Demetra
Ho, Lewis
Dafoe, Allan
Computers and Society
Artificial Intelligence
Frontier AI governance often centres on the model-level governance paradigm, which assumes that a model's capability profile is primarily a function of the compute and data used during training. This position paper argues that model-level governance becomes less effective when capability progress is increasingly driven by "non-model gains"--improvements that are independent from advances in the base model. We formalise the concept of non-model gains and provide a taxonomy of three distinct vectors of capability gain: inference gain (scaling compute at test-time), systems gain (post-training enhancements such as scaffolds), and asset gain (enhancing a model with restricted assets). We demonstrate how these vectors--alongside potential future impacts from embodiment, continual learning, and AI diffusion--may undermine risk management strategies that hinge mostly on pre-deployment evaluation and mitigation. We provide an overview of governance approaches that go beyond the model level: system, entity, agent, and cloud governance. Finally, we emphasise the importance of societal resilience as a complement to these governance layers.
title Comprehensive AI governance requires addressing non-model gains
topic Computers and Society
Artificial Intelligence
url https://arxiv.org/abs/2606.00047