Governments were urged to tighten controls on commercial gene synthesis to prevent vendors being exploited to build biological weapons, because rapid advances in artificial intelligence are eroding the expert knowledge that long kept such threats out of reach. The public letter was signed by chief executives and founders from major AI organisations, including Sam Altman of OpenAI, Dario Amodei of Anthropic, Demis Hassabis of Google DeepMind and Mustafa Suleyman of Microsoft AI, and it called for laws requiring companies that sell synthetic genetic material to screen customers and orders. The organisers, the Institute for Progress and the Foundation for American Innovation, argued that AI tools can now help users identify suppliers and modify orders so as to evade current checks. This next concrete step is for lawmakers to decide whether to adopt statutory screening obligations for suppliers of synthetic DNA and RNA.

Governments should force routine vetting of gene customers because AI can lower the technical barrier to misuse, the letter said, and voluntary industry checks won't be enough. It set out a precise policy fix: require commercial gene synthesis suppliers to screen buyers and to use automated screening of sequence orders so that companies can't be easily bypassed by malicious users guided by AI.

The public signatories included senior figures from across the AI industry. Sam Altman, chief executive of OpenAI, Dario Amodei of Anthropic, Demis Hassabis of Google DeepMind and Mustafa Suleyman of Microsoft AI were named among chief executives and founders who appended their names. The letter also drew scientists, national security experts and executives from gene synthesis firms, including representatives from Twist Bioscience and Ansa Biotechnologies, which are members of the International Gene Synthesis Consortium.

The organisers framed the plea as a response to a changing technical baseline. The Institute for Progress and the Foundation for American Innovation said the speed of AI development creates a "real possibility" that the expertise historically needed to construct biological weapons will meaningfully erode. Signers stressed that DNA synthesis is now automated and cheaper than in past decades, and that some providers don't screen all orders or enforce buyer qualifications.

David Relman, a microbiologist and biosecurity expert at Stanford University who signed the letter, put the risk in practical terms. "AI tools enable a user to very quickly identify where to turn to order sequences that won't be subject to screening," he said. "Appropriately prompted models can also suggest how to alter an order so that even screened systems may fail to detect dangerous intent." The letter used that sort of operational detail to argue why voluntary commitments aren't enough.

Technical precedent and industry practice

The signers cited concrete precedents to show the gap between past and present risk. They pointed to a 2017 experiment in Canada in which researchers used roughly $100,000 of mail-order DNA to reconstitute the extinct horsepox virus, an episode critics said showed how existing ordering channels might be abused to recreate smallpox-like agents.

Mastheads

That case underpins the letter's claim that cheaper, automated synthesis can be misused through ordinary commercial channels.

Industry screening isn't new. The International Gene Synthesis Consortium has promoted voluntary screening practices since 2009, and several providers already vet customers and sequences. But the letter noted that not all firms use the same checks. Some suppliers, the signers said, don't screen orders or enforce buyer qualifications consistently, leaving gaps that could be exploited.

Independent commentary cited by the organisers strengthens the technical argument. Forbes published analysis pointing out that frontier AI models increasingly outperform expert virologists on complex benchmarks and that in 2025 multiple AI developers imposed additional safeguards after internal testing found models could meaningfully aid novices trying to design or modify pathogens. Those developments are central to the signers' contention that AI is lowering the knowledge barrier for misuse.

The policy prescription in the letter is specific rather than abstract. It asks lawmakers to require gene-synthesis vendors to screen customers and sequence orders, and to strengthen tracking and oversight so companies can't be easily bypassed by malicious users guided by AI. The organisers described these measures as a complement to existing industry screening software that flags "sequences of concern," not as a replacement for scientific or technical safeguards.

The letter was organised publicly by the Institute for Progress and the Foundation for American Innovation, signalling an effort to move the debate from voluntary norms to statutory obligations. Signers included executives from gene-synthesis firms, which matters because those companies already operate the supply chains at issue. Twist Bioscience and Ansa Biotechnologies, both named in the letter, are members of the International Gene Synthesis Consortium and thus already participate in voluntary screening frameworks.

But the signers argued that voluntary arrangements can be fragile in the face of tools that help users find suppliers willing to ship unchecked sequences or that suggest simple modifications to evade detection. That mix of cheaper synthesis, automation and capable AI is the proximate cause the letter identifies for urging new legal rules.

The public text didn't attach a timetable or specific legislative language. Instead it explicitly urged members of the US Congress to enact the proposed legal requirements for screening and oversight, thereby handing the ball to lawmakers to determine whether and how to convert the signers' proposals into statute.

Related Articles

The letter asked members of the US Congress to convert the signers' proposed screening and oversight rules into law. It contains no timetable, so the next concrete decision now rests with legislators.

This article was created with AI assistance.