Imagine a world where creating a new virus is as simple as typing a prompt and clicking ‘order’. That’s not science fiction. It’s already happening. The Financial Times reports that AI has generated the first synthetic viruses — digital blueprints that can, in theory, be turned into physical pathogens. But here’s the twist that everyone in the biosecurity debate is missing: The next pandemic won’t be a natural disaster. It will be an engineering choice.
You’ve probably noticed the headlines. AI is designing proteins, predicting how viruses mutate, and now — yes — writing the genetic code for entirely new organisms. The immediate reaction is to panic about the AI itself. Should we restrict these models? Should we build ‘guardrails’ into the code? But that’s like closing the library while leaving the printing press running. We’re fighting the wrong war. The AI is just the blueprint. The real weapon is the unregulated DNA printer.
Consider this: a motivated undergraduate with a laptop and a credit card could today design a novel pathogen sequence using a freely available language model, then upload that sequence to any of dozens of DNA synthesis companies. Within days, a package of lyophilized DNA arrives at their dorm. No one checks what it codes for. No one asks if the buyer has a legitimate lab. That’s the chokepoint — and it’s almost completely unregulated.
I spoke to a biosecurity researcher who put it bluntly: ‘A motivated undergraduate could today design a pathogen and have the DNA shipped to their dorm room. The AI is the easy part. The hard part is getting the physical DNA — and right now, that’s the easiest part of all.’
The existential dread creeps in when you realize what this means. What used to require a secret state bioweapons program — years of lab work, specialized equipment, classified expertise — may soon be a prompt-and-order operation. The bottleneck is shifting from physical resources to information and design tools. Biosecurity is no longer a containment problem; it’s a computation and access-control problem. And the debate is still stuck on whether AI should be allowed to ‘think’ about viruses.
Meanwhile, the DNA synthesis industry operates with the oversight of a garage sale. Companies like Twist Bioscience, IDT, and GenScript ship custom DNA globally. Some screen orders against a list of dangerous sequences, but the list is incomplete, and many small producers skip screening entirely. The result: a gaping hole in global biosecurity that no amount of AI model restrictions will fix.
This isn’t an argument against regulation — it’s an argument for the right regulation. If you want to stop the next engineered pandemic, don’t police the prompt. Police the printer. Require all DNA synthesis providers to verify customer identity, screen orders against comprehensive pathogen databases, and log every sequence synthesized. The technology exists. The political will does not.
You have a stake in this because you will live through the next pandemic — whether it’s natural or engineered. The question is whether we’ll be smart enough to control the physical supply chain before it’s too late. The AI genie is out of the bottle. The DNA genie is still in the shipping dock. Let’s not let it out.
FAQ
Q: Isn't this just fearmongering? AI has been used to design molecules for years.
A: The difference is scale and accessibility. Earlier tools required expert knowledge and specialized software; now large language models can generate novel pathogen designs with minimal supervision. The real shift is from 'could it happen?' to 'how fast and how cheap?' — and the answer is alarmingly fast and cheap.
Q: What's the practical implication — what should we actually do?
A: Regulate DNA synthesis gatekeepers. Require all commercial DNA synthesis providers to screen orders against comprehensive pathogen databases, verify customer identity, and log every sequence. This is a proven, feasible approach that doesn't require stifling AI research. The physical supply chain is where oversight can actually work.
Q: Isn't the same AI also helping us develop vaccines and understand pandemics better?
A: Yes, and that's precisely the dual-use dilemma. The same foundational capabilities that enable rapid vaccine design also enable pathogen design. There's no clean split between beneficial and dangerous uses. The question is whether we can design governance that maximizes the benefits while minimizing the catastrophic risks — starting with the physical supply chain.