Dual Use in AI-Driven Biotechnology: How Should We Govern It?

This technical commentary discusses the dual-use implications of artificial intelligence in biotechnology, with particular attention to bioinformatics, smart biomanufacturing, and AI-enabled research infrastructure. The article examines how AI systems can accelerate biological research by supporting genome interpretation, protein-function prediction, literature mining, strain discovery, process monitoring, and Design-Build-Test-Learn workflows. At the same time, it argues that the same capabilities can lower barriers to misuse when biological knowledge, datasets, computational tools, and automation layers become easier to access and combine. The central argument is that AI in biotechnology should not be treated simply as a productivity tool or as an external chatbot. As AI systems become connected to databases, code execution, laboratory workflows, and decision-support systems, they become part of the laboratory stack. This requires a governance model that balances open science, legitimate research acceleration, and biosecurity risk management. The article proposes biosecurity-by-design as a practical compromise, extending Safe-by-Design and Safe-and-Sustainable-by-Design principles to AI-enabled biotechnology. It discusses tiered access, traffic-light risk classification, tool-level restrictions, auditable logging, human oversight, and continuous red teaming as components of a governed AI-biotechnology infrastructure. This Zenodo record archives the article originally published on the Bio-Mosaic website.