The development of large language models has reached a point where they can perform complex biological tasks, such as literature synthesis and experimental data interpretation, with a level of proficiency comparable to experienced human biologists. This has significant implications for biosecurity, as these AI agents can now conduct in silico biology tasks that were previously exclusive to human experts. The introduction of the ABC-Bench benchmark aims to assess the biosecurity capabilities of these AI models, evaluating their potential to facilitate malicious activities, such as the design of biological agents1. The benchmark's focus on agentic bio-capabilities highlights the need for a comprehensive understanding of the risks associated with AI-driven biological research. As AI continues to advance in the biological sciences, the potential consequences for policy, security, and workforce dynamics become increasingly pressing, making it essential for practitioners to stay informed about the latest developments in this field.