Artificial intelligence is altering the risk landscape around two of the world's most dangerous technologies: biological agents and nuclear weapons. According to the Bulletin of the Atomic Scientists, AI is changing the nature of these biological and nuclear risks, and the systems meant to govern them need to change in step.
The core argument is one of mismatch. Rules, treaties, and oversight bodies built for an earlier era were not designed with rapidly advancing AI in mind. As the Bulletin frames it, if the underlying risks are shifting because of AI, then governance cannot stand still — it must adapt to keep pace.
The Bulletin of the Atomic Scientists is a long-standing voice on existential threats, best known for its assessments of nuclear and emerging-technology dangers. Its call here is directed at how societies oversee and contain misuse, rather than at any single product or company.
The source item does not lay out specific new regulations, technical mechanisms, or numerical estimates. What it establishes is the direction of the problem: AI is a force multiplier and a wildcard layered on top of fields where the stakes are already catastrophic, and the institutions responsible for safety have to account for that shift.
Why it matters: biological and nuclear domains leave little room for error, so a gap between fast-moving AI capabilities and slow-moving governance is exactly the kind of blind spot that experts warn can turn manageable risks into unmanaged ones.