This development deserves a serious biosecurity warning. The immediate experiment involved bacteriophages, not human pathogens. But the important threshold is that AI can now help design complete genomes that become functional biological entities after synthesis. The danger is what increasingly capable systems may eventually be able to design: self-replicating pathogens, mirror microorganisms, novel toxin-producing systems, transmissible prion-like agents, or entirely new biological threats we have not yet classified. Mirror life is especially concerning because researchers have already warned that self-replicating mirror organisms could evade important biological defenses and create potentially irreversible health and ecological risks. We should not wait until one of these capabilities is demonstrated. Biosecurity needs independent safeguards across the entire chain: AI capability evaluation, synthesis screening, laboratory oversight, containment, auditing, and international coordination. The governing principle should be simple: No person, laboratory, company, or AI system should have an unchecked end-to-end pathway from unrestricted biological design to synthesis, testing, optimization, and deployment of potentially catastrophic biological agents. The benefits of AI biotechnology are enormous. That is exactly why the catastrophic-use boundary needs to be established before the technology reaches it.