Research
We have described a regulatory regime that could be imposed on biological research, manufacturing and agriculture in the United States following a triggering incident caused or assisted by artificial intelligence. The regime is strict and has four parts. It would make unlicensed operation a criminal matter, would give the authorities a detailed picture of the operation of the biological enterprise, would place particular lines of work before an independent panel, and would do so through a body we have called the Biological Capability Review Group.
The finding on which the plan rests is bibliometric. Examining the published methods by which virus-length recombinant DNA constructions have been built, we find that advanced DNA polymerases are key reagents, and that controls placed on those reagents would therefore be narrow rather than broad. The analysis indicates which materials would repay control and which would not. The paper is Brent and Colman-Lerner (2026), posted at Preprints.org.
Two pieces of work follow. The first is a list, comparable in function to the control lists maintained by the Australia Group, of the reagents, materials, equipment and services that ought to be subject to regulation after a sufficiently serious triggering event. The names on such a list can be public, and we intend that they should be.
The second is the independent panel itself, which we have called the Risk Benefit Review Board, and whose members are to be distinguished scientists who do not work on infectious disease or public health, together with chemists and experimental physicists.
We also develop technical measures intended for use in emergencies, as distinct from the regulatory measures described above. These are prepared in advance, tested, and discussed with the parties who would need to act. We do not describe them here.
The purpose of this work is preparation rather than prevention. The regime is not imposed now, and we do not propose that it should be. It is written and analyzed so that, if a triggering event occurs, the response available to those who must respond is specific and immediate rather than improvised under conditions of alarm. A regime drafted in haste after such an event would be likely to control the wrong things, and to control far more of them than necessary.