Some of the biggest names in artificial intelligence just dropped a bombshell on Capitol Hill, and it has nothing to do with chatbots or image generators. Google DeepMind’s Demis Hassabis, OpenAI’s Sam Altman, Anthropic’s Dario Amodei, and Microsoft AI’s Mustafa Suleyman signed a public letter asking Congress to pass laws that would make it harder for bad actors to use AI to develop biological weapons.
It’s a striking moment. These are the same companies building the most powerful AI systems on the planet, and they’re essentially telling lawmakers, “Hey, we can see the edge of a cliff coming, and we’d rather not drive off it.”
The Gene Synthesis Wild West
The letter focuses on gene synthesis companies, the businesses that use commercial synthesizers to print custom DNA and RNA sequences. These sequences power scientific research, drug development, and diagnostics. Sounds useful, right? It absolutely is. The problem is that not every provider screens their customers or the orders they receive.
Here’s where it gets uncomfortable. In 2017, Canadian researchers used roughly $100,000 worth of mail-order DNA to reconstitute the horsepox virus, an extinct relative of smallpox. The scientific community raised alarms. The same methodology could, in theory, be used to construct smallpox itself. And gene synthesis has only gotten cheaper since then.
Now layer AI on top of this. Large language models can already help someone identify where to order sequences that won’t trigger screening software. Ask the right prompts, and these models can even suggest how to modify an order so it slips past existing safeguards. That’s not science fiction. That’s what a Stanford biosecurity expert told the letter’s organizers.
The Industry Is Asking to Be Regulated
What makes this letter noteworthy is who’s signing it. We’re not just talking about AI executives. The signatories include gene synthesis companies like Twist Bioscience and Ansa Biotechnologies, both members of the International Gene Synthesis Consortium that formed in 2009 to implement voluntary screening practices.
James Diggans, vice president of policy and biosecurity at Twist Bioscience, put it simply: if you have the technology to synthesize DNA, you have a responsibility to understand what you’re making and who you’re making it for. That’s a remarkable stance from a company that profits from selling exactly that technology.
The industry has been pushing for formal rules for years. Federal guidelines introduced during the Biden administration already require scientists receiving federal funding to order synthetic gene sequences only from providers that screen purchases. A bipartisan Senate bill introduced earlier this year would expand that requirement to all gene synthesis providers operating in the US.
But Screening Isn’t Enough
Here’s the uncomfortable truth that even the letter’s signatories acknowledge: screening tools aren’t perfect. Last year, Microsoft researchers published a study showing that AI protein design tools could generate potentially dangerous gene sequences that slipped past companies’ screening software. The models suggested new protein sequences with structures similar to known dangerous ones.
There’s a perverse dynamic at play here. The same tools designed to accelerate beneficial biological research can be weaponized with the right prompts. And once AI assistance enters the picture, the knowledge barriers that have historically prevented bad actors from creating biological weapons start to look a lot lower.
Geoff Ralston, former president of Y Combinator and a partner at the Safe AI Fund, thinks AI labs with biology models should do their own screening of users. It’s a logical extension: if your model can help design dangerous proteins, maybe you have a duty to stop certain queries from ever reaching it.
So Where Does This Leave Us?
The honest answer is that nobody really knows how to solve this problem perfectly. Screening orders is necessary but insufficient. Regulating AI models is politically fraught and technically challenging. And the underlying technology is only going to get more powerful and more accessible.
What’s clear is that the companies building these systems see the risk clearly enough to go public with it. That’s either a sign of genuine concern or a strategic move to get ahead of regulations they know are coming. Probably a bit of both.
What concerns me most is the pace mismatch. Biology is no longer the slow-moving science it was even a decade ago. AI is accelerating at a breakneck speed. And regulatory frameworks, by their nature, move at a glacial pace. The gap between what’s possible and what’s governed is widening by the month.
We’re entering territory where a single individual with the right knowledge and access could potentially cause catastrophic harm. That’s a profound shift in the biosecurity landscape, and it’s happening with barely a whisper of public attention.
