Securing Dual-Use Pathogen Data of Concern

TL;DR

The paper addresses securing dual-use pathogen data used to train biological AI by proposing a five-tier Biosecurity Data Level (BDL) framework that classifies data by its potential to enable concerning capabilities, with escalating restrictions. It couples this taxonomy with technical data-access controls and trusted research environments to enable legitimate research while mitigating misuse, and introduces an independent Pathogen Data Board to govern data classification and oversight. The work emphasizes empirical validation of data-type-to-capability links, practical TRE-based implementations, and international alignment to balance openness with safety in a globally connected research landscape. Its significance lies in providing a structured, governance-driven approach to preemptively regulate high-risk data before it contributes to harmful AI capabilities, while preserving the majority of benign data for beneficial scientific progress.

Abstract

Training data is an essential input into creating competent artificial intelligence (AI) models. AI models for biology are trained on large volumes of data, including data related to biological sequences, structures, images, and functions. The type of data used to train a model is intimately tied to the capabilities it ultimately possesses--including those of biosecurity concern. For this reason, an international group of more than 100 researchers at the recent 50th anniversary Asilomar Conference endorsed data controls to prevent the use of AI for harmful applications such as bioweapons development. To help design such controls, we introduce a five-tier Biosecurity Data Level (BDL) framework for categorizing pathogen data. Each level contains specific data types, based on their expected ability to contribute to capabilities of concern when used to train AI models. For each BDL tier, we propose technical restrictions appropriate to its level of risk. Finally, we outline a novel governance framework for newly created dual-use pathogen data. In a world with widely accessible computational and coding resources, data controls may be among the most high-leverage interventions available to reduce the proliferation of concerning biological AI capabilities.