Regulatory & legal
Export control and dual-use compliance
How dual-use biological controls are constructed, why control by organism name breaks once synthesis is de novo, and the functional-equivalence problem that limits sequence screening.
Export control over biological material was built on an assumption that held for forty years: dangerous biology travels as a physical thing, and a physical thing can be named. The Australia Group, formed in 1985, maintains common control lists that enumerate human, animal and plant pathogens and toxins, plus the equipment — fermenters above a stated volume, certain aerosol chambers, containment gear — through which national regimes implement controls. These lists are transposed into the US Commerce Control List under Export Control Classification Numbers 1C351, 1C352 and 1C353, and into the EU dual-use regime, Regulation (EU) 2021/821. A shipment is controlled because of what it is declared to be, and the enforcement act is a licence decision at a border.
What makes a thing controllable
Control by enumeration works when three conditions hold: the controlled item is discrete, it is identifiable by a name that a customs declaration can carry, and possessing the name is materially harder than possessing the list. Equipment satisfies all three. A fermenter has a serial number and a manufacturer. A culture of a listed organism arrives in a vial with a taxonomic label.
De novo synthesis breaks the first two conditions at once. When a genome is assembled from ordered oligonucleotides, nothing crossing a border is the organism; what crosses is sequence information, and information can be transmitted without a declaration at all. Control lists therefore added genetic elements — nucleic acids coding for a controlled agent’s pathogenicity or toxin genes — but this shifts the enforcement problem from checking a label to reading a sequence.
Why screening is a matching problem, and where matching fails
Providers screen orders against sequences of concern. The International Gene Synthesis Consortium’s Harmonized Screening Protocol is the industry’s own version; in the United States, the 2010 HHS screening framework for suppliers of synthetic double-stranded DNA and, later, the 2024 OSTP Framework for Nucleic Acid Synthesis Screening set out federal expectations, the latter reaching customers and equipment as well as sequences.
The technical limit is functional equivalence. The genetic code is degenerate: a protein of a few hundred residues has an astronomically large number of nucleotide encodings, so any exact-match test at the DNA level is defeated by silent recoding that changes nothing about the product. Translated homology search closes that gap and is what serious screening uses — but it inherits two failures of its own. Homologous sequences are shared with organisms nobody controls, so a strict threshold generates false positives at a rate that makes screening unusable; ricin is a controlled toxin while its source plant is an ornamental. And proteins that do the same thing with low sequence identity are invisible to homology altogether, because homology detects shared ancestry, not shared function.
No screening method available today tests function. It tests resemblance to things already known to be dangerous, which means the regime’s coverage is bounded by its reference database rather than by biology. Benchtop synthesisers sharpen this: they move the screening decision from a small number of auditable providers to the instrument itself, where the assurance is a software control rather than a contractual one.