AI-enabled biological research is becoming a connected system of biological data, AI models and agents, computational tools, human researchers, automated laboratories, and physical biological processes. Yet many approaches to AI-biosecurity still evaluate individual models or capabilities rather than asking how knowledge, capability, risk, and resilience emerge across the relationships among them.
In this talk, I introduce relational biosecurity and an emerging research program aimed at making AI-enabled biological R&D observable. I distinguish development observability (whether a system is functioning safely and reliably) from research observability (how the researcher and the system came to know what they think they know). I ask what information, provenance, uncertainty, control, and feedback must remain visible as biological information moves across data, models, tools, people, experiments, and institutions.
I will describe an empirical approach based on mapping relationships, structured observation of existing AI-bio workflows, and perturbation of selected relationships to identify a minimum observable core that can support comparison across very different system types. The longer-term goal is to translate those findings into reporting practices, metadata, interfaces, and infrastructure that make observability increasingly native to scientific workflows.
The broader aim is to connect beneficial biological research with security research through a shared empirical substrate: improving reproducibility and scientific learning while strengthening detection, attribution, recovery, and public-interest oversight of increasingly capable AI-enabled biological systems.
Michelle Holko, Ph.D., PMP, is a scientist and strategist working at the intersection of biology, artificial intelligence, cybersecurity, and national security. She has developed and led biotechnology and biosecurity programs across DARPA, BARDA, CISA, NIH, Google, and the private sector, spanning biological data infrastructure, AI-enabled biology, infectious-disease detection, and emerging-technology risk. She is currently Principal Scientific Advisor at Computercraft supporting NIH/NLM/NCBI, an Adjunct Senior Fellow at the Center for a New American Security, and Principal Investigator at ICSI Berkeley. Her recent work focuses on relational biosecurity and the security implications of increasingly connected biological data, AI, computational, human, and physical systems.