treats living systems as components that can be designed and assembled.
The applications are unglamorous and substantial: , fuels, materials and vaccines produced in engineered organisms.
The risk that dominates policy discussion is deliberate rather than accident.
The concern is reasonable and the standard framing of it is imprecise.
Designing a dangerous organism has never been the difficult step.
Producing it, stabilising it and delivering it are difficult, and they require tacit skills that a published does not confer.
Policy that concentrates on information control therefore addresses the least binding constraint.
at the point of synthesis is the measure that fits the actual bottleneck.
Companies that DNA to order can check s against a list of agents of concern and verify the customer.
Most of the established firms do so voluntarily, and the practice is not universal.
A requirement attached to the purchase of synthesis equipment reaches the firms that a voluntary code does not.
rs complicate this, because they move the capability into any laboratory that can afford one.
Building the screening into the device, so that it refuses certain s, is the proposal under discussion.
It is technically feasible and creates a governance problem of its own.
Somebody must maintain the list, and that list is then a piece of security infrastructure held by a private firm or by one state.
The wider lesson is that resembles cybersecurity more than it resembles arms control.
The relevant capability is distributed, mostly civilian and improving for reasons nobody wishes to reverse.
Governance that assumes a small number of state actors will therefore describe a world that no longer exists.