US-China Biorisk Cooperation: Yes, It’s Possible
ChinaTalk Nick Corvino
US and China could both start screening DNA orders for risky sequences. It’s a rare AI safety move that’s practical, narrow, and doesn’t slow biotech much.
Based on reporting by ChinaTalk, Nick Corvino — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
A strange thing happened in a June open letter: Sam Altman, Dario Amodei, Demis Hassabis, and a mix of AI and biosecurity voices from left and right landed on the same concrete ask. They want the US to require nucleic acid synthesis companies to screen DNA and RNA orders before shipping them. That sounds technical, but the logic is simple. AI is making it easier to design dangerous biology, and synthesis firms sit right where digital instructions become physical material.
Washington has already flirted with this. The Trump administration scrapped a Biden-era screening framework last year, said it would replace it with something stronger, and then mostly moved on. In Congress, there is still a bill from Cotton and Klobuchar, plus a narrower House bill from Pfluger and Houlahan. The case for action is not abstract fear-mongering either. It is the ugly reality that one underregulated supplier can turn into the loophole that ruins the whole system.
The bigger point is that this is not just an American problem. China accounts for about 34% of the world’s DNA synthesis providers, and its AI ecosystem is more open-source, which makes model-level controls harder to rely on. If orders can simply be routed through looser jurisdictions, a US-only rule leaves a hole. That is why the International Biosecurity and Biosafety Initiative for Science has been mapping the global supply chain, with more than 700 providers in 81 countries, and why this belongs on the agenda for upcoming US-China AI talks.
Synthesis screening is appealing because it is narrow and verifiable. Providers compare custom DNA and RNA orders against databases of dangerous pathogens and toxins, check the customer, and decide whether to fill the request, hold it, or reject it. The point is not to police every lab experiment. It is to keep someone from quietly ordering the pieces of a toxin or pathogen and having them show up at the door with no scrutiny.
And the cost argument is much weaker than the usual biotech panic suggests. Ordering DNA takes days; making a drug takes years. The hard parts are downstream: failed constructs, cell work, animal studies, clinical trials, manufacturing, regulators. Firms representing roughly 80% of global synthesis capacity already screen voluntarily, and the software exists. Making it mandatory mostly forces the laggards to catch up.
That is why this looks like one of the few real openings for US-China cooperation on AI safety. Both governments have the same incentive here: don’t let a Covid-like disaster come out of your own backyard. If they can each tighten their own supply chain without waiting for the other side to blink first, that is not grand diplomacy. It is basic competence with a side of survival instinct.
My take — AI-written commentary, not fact-checked reporting
This is the rare AI safety idea that doesn’t smell like a regulation hobbyhorse. Screen the DNA orders, not the moonlight; leave the model drama to the pundits and fix the chokepoint that actually matters. The broader lesson is simple: when open-source AI and sloppy bio supply chains meet, the bill comes due in the physical world, not on a benchmark chart.
Read more about this at: ChinaTalk