Generative AI in microbial evolution and resistance: toward robust, explainable, and equitable predictions

Antimicrobial resistance (AMR) is one of the most urgent challenges in modern microbiology, both an evolutionary inevitability and a global health crisis shaped by clinical practices, ecological disruption, and social inequities. Generative artificial intelligence (AI) and large language models (LLMs) present new opportunities to anticipate resistance pathways, design novel antimicrobial agents, and guide interventions that are informed by evolutionary dynamics. Their successful integration, however, depends on addressing three fundamental imperatives. The first is evolutionary robustness, requiring models that incorporate mutation, horizontal gene transfer, and adaptive landscapes to move beyond retrospective classification toward predictive evolutionary inference. The second is explainability and biosafety, which demand interpretable and biologically credible outputs that clinicians, microbiologists, and policymakers can trust, while safeguarding against dual use risks. The third is data equity, which calls for strategies that mitigate structural biases in global microbial datasets and ensure that predictive systems serve the populations most affected by AMR. This Perspective advances the view that generative AI must be conceived as a transformative epistemic infrastructure that is evolution aware, transparent, and globally inclusive, capable of supporting sustainable drug discovery, adaptive surveillance, and equitable microbiological futures.