Regulatory and governance frameworks are lagging behind AI-enabled biology, with existing rules not fully addressing AI-driven automation and the need for coordinated action or new managed-access models.

Current regulations lag behind AI-driven biology, as some firms pursue voluntary safety measures while policy efforts explore managed access and enhanced screening.

GPT-5 has autonomously designed and ran tens of thousands of biological experiments, achieving rapid iterations from study design to data feedback in a closed loop, with humans setting the objectives.

Similar systems autonomously design and execute large numbers of biological experiments, with human goals and machines handling execution and data-driven design refinement.

AI-driven protein design uses protein language models to predict mutations and craft new proteins, speeding drug development and vaccine responses when paired with automated laboratories.

This AI-enabled design and engineering creates fast iterative loops that accelerate drug development and responses to emerging infections, contingent on automated lab execution.

Experts emphasize that AI can accelerate science with proper controls, but insufficient oversight risks serious biosecurity harms as capabilities advance.

Studies show mixed results on novices performing biosecurity tasks with AI assistance, underscoring containment challenges as automation lowers labor barriers in the lab.

Evidence indicates AI assistance yields mixed outcomes for novices in performing bio-lab tasks, highlighting the need for safeguards as automation expands beyond trained personnel.

Policy recommendations call for improved DNA synthesis screening, pre-release model risk evaluations, and governance of biological data through proactive international and national frameworks to balance risk and innovation.

Experts urge coordinated action to strengthen screening, model evaluations, and data governance to mitigate biosecurity risks without stifling scientific advancement.

The dual-use risk is significant: AI tools can both optimize viral spread and enable risky biological work, exposing gaps in safety filters and governance.