[Paper Review] AI Model Registries: A Foundational Tool for AI Governance

Elliot McKernon, Gwyn Glasser+2 more|arXiv (Cornell University)|Oct 12, 2024

Scientific Computing and Data ManagementDecision Sciences3 citations

TL;DR

This paper proposes national registries for frontier AI models as a foundational tool for AI governance, enabling governmental oversight through standardized data collection on model architecture, size, compute, and training data, while minimizing regulatory burden via injunctive enforcement and financial penalties. The framework enhances AI safety and innovation parity without compromising intellectual property.

ABSTRACT

In this report, we propose the implementation of national registries for frontier AI models as a foundational tool for AI governance. We explore the rationale, design, and implementation of such registries, drawing on comparisons with registries in analogous industries to make recommendations for a registry that is efficient, unintrusive, and which will bring AI governance closer to parity with the governmental insight into other high-impact industries. We explore key information that should be collected, including model architecture, model size, compute and data used during training, and we survey the viability and utility of evaluations developed specifically for AI. Our proposal is designed to provide governmental insight and enhance AI safety while fostering innovation and minimizing the regulatory burden on developers. By providing a framework that respects intellectual property concerns and safeguards sensitive information, this registry approach supports responsible AI development without impeding progress. We propose that timely and accurate registration should be encouraged primarily through injunctive action, by requiring third parties to use only registered models, and secondarily through direct financial penalties for non-compliance. By providing a comprehensive framework for AI model registries, we aim to support policymakers in developing foundational governance structures to monitor and mitigate risks associated with advanced AI systems.

Motivation & Objective

• To address the lack of governmental insight into high-impact AI systems, mirroring oversight in regulated industries like aviation and pharmaceuticals.
• To reduce risks associated with frontier AI by creating a centralized, transparent, and efficient mechanism for monitoring model development.
• To support responsible AI innovation by minimizing compliance burden while ensuring accountability.
• To balance transparency with protection of intellectual property and sensitive information in model registration.
• To establish a governance framework that enables timely, accurate, and enforceable oversight of advanced AI systems.

Proposed method

• Design a national registry system for frontier AI models, modeled on existing registries in high-risk industries such as aviation and pharmaceuticals.
• Define core data elements to be collected: model architecture, model size, compute resources, training data, and evaluation results.
• Introduce a tiered enforcement mechanism: primary reliance on injunctive action requiring third parties to use only registered models, and secondary financial penalties for non-compliance.
• Ensure data privacy and IP protection through mechanisms like redaction, access controls, and limited public disclosure of sensitive information.
• Propose evaluation frameworks tailored to AI systems to assess safety, robustness, and alignment, enhancing the utility of registry data.
• Integrate the registry into broader AI governance ecosystems to support policy development, risk monitoring, and incident response.

Experimental results

Research questions

• RQ1 How can governments achieve effective oversight of frontier AI models without impeding innovation?
• RQ2 What core information should be collected in a model registry to enable meaningful risk assessment and regulatory insight?
• RQ3 How can intellectual property and sensitive technical details be protected within a public-facing or semi-public registry?
• RQ4 What enforcement mechanisms are most effective in ensuring compliance with model registration requirements?
• RQ5 To what extent can existing regulatory models from other high-impact industries be adapted to AI model governance?

Key findings

• National AI model registries can provide critical governmental insight into frontier AI systems, comparable to oversight in aviation and pharmaceuticals.
• Collecting standardized data—such as model size, compute, and training data—enables risk monitoring and supports evidence-based policy.
• Injunctive enforcement, requiring third parties to use only registered models, is a more effective compliance mechanism than financial penalties alone.
• The registry framework can be designed to respect intellectual property and safeguard sensitive information through controlled access and data minimization.
• Evaluation tools tailored for AI can significantly enhance the utility and reliability of registry data for governance purposes.
• A well-designed registry supports both safety and innovation by reducing uncertainty and enabling proactive risk mitigation.