13BiosecurityBelow the top 20
Protein-design models for novel toxins
CapabilityDomain knowledgeImpact domainBoth
Statement (NASA form)
Given that open and commercial protein models already propose folds no paper has named, there is a possibility of a designed protein with toxin-like function that screening does not recognise resulting in a biological hazard that never existed in a natural genome.
Likelihood
2Remote
Consequence
5Catastrophic
Urgency
3Priority
Sequence screens fail when the threat is a shape, not a known string.
Composite 13 = 2×5 + 3
Applicable mitigations
Controls · Mandatory, model-aware synthesis screening · Restore independent biosecurity staff at frontier labs · Know-your-customer for high-risk biological queries · Capability thresholds that cannot be open-released · Know-your-customer on high-end fine-tune clusters · Pair open release with synthesis and exploit screening upgrades