Biosecurity frameworks developed alongside molecular biology's incremental advances — each new capability arrived with enough lead time that oversight bodies could update their reference classes. Generative AI protein design has broken that rhythm. The Spiez Laboratory mini-review published in Frontiers in Microbiology establishes that biological security risks from novel protein design now include molecules that have no evolutionary precedent, which means the standard audit trail — trace the sequence, assess its fitness history, compare against known threats — terminates at the point of AI generation. The reference class simply does…