When Your Fiercest Rivals Agree on Something, Congress Should Listen

OpenAI, Anthropic, Google DeepMind, and Microsoft AI just issued the most significant collective safety warning the AI industry has ever produced — and the ask is surprisingly specific.

Sam Altman and Dario Amodei have spent years disagreeing in public. Their companies are building competing frontier models, competing for the same talent, and competing for the same future. On June 4, 2026, they signed the same letter.

So did Demis Hassabis (Google DeepMind), Mustafa Suleyman (Microsoft AI), and Alexandr Wang (Meta’s chief AI officer). Scientists like Nobel laureate David Baker added their names too. The combined letter — hosted at screendna.org and backed by the Institute for Progress and the Foundation for American Innovation — represents the most unified public statement the AI industry has ever made on a safety issue.

The message: AI is eroding the knowledge barriers that have historically kept bioweapons out of reach. And Congress needs to act before the window closes.

What the letter actually asks for

This is not a vague “be responsible about AI” plea. The letter has a specific, narrow target: companies that sell synthetic DNA and RNA — biological building blocks that can be ordered online and delivered like lab supplies.

Right now, many of these companies screen their orders voluntarily. Some vetting happens. But there are gaps, the standards are uneven, and “voluntary” has never been a reliable foundation for a threat this serious.

The letter asks Congress to mandate two things:

1. Customer and order screening. Every company selling synthetic nucleic acids would be required to verify who is buying and flag sequences that match dangerous biological patterns before the material ships.

2. Mandatory recordkeeping. Detailed transaction records, so that investigators can trace multi-part orders that each look innocent alone but together could enable the construction of a dangerous pathogen.

Crucially, the authors are not calling for a ban on synthetic biology. They are calling for smarter gatekeeping at a specific chokepoint — one that is already commercially viable to implement. The companies that manufacture synthetic DNA, including Twist Bioscience and Ansa Biotechnologies, also signed the letter. They are not being regulated over their objections. They are asking to be regulated.

Why AI changes the calculus

Synthetic biology has always carried dual-use risk. For decades, biosecurity researchers have pushed for tighter controls on who can order what genetic sequences and from whom. Progress has been slow but real. Voluntary screening agreements have helped. The knowledge required to actually design and produce a dangerous pathogen has been a significant barrier.

AI is attacking that barrier.

According to the letter’s signatories, frontier AI models now outperform PhD-level virologists on many technical lab questions. That does not mean anyone with an API key can now synthesize a pandemic. The physical steps still require equipment, expertise, and controlled materials. But the knowledge layer — the part that made bioweapons historically hard to design for non-specialists — is thinning.

The concern is not today. It is the slope. AI capability is improving faster than biosecurity infrastructure is adapting. Waiting until the gap is obvious is waiting too long.

As the letter states: “AI systems are improving rapidly, and alongside incredible benefits to science and medicine, there is a real possibility that the knowledge barriers which have historically prevented bad actors from obtaining biological weapons will meaningfully erode.”

The bill that already exists

Congress has not been entirely asleep on this. In February, Sen. Tom Cotton (R-Ark.) and Sen. Amy Klobuchar (D-Minn.) introduced the Biosecurity Modernization and Innovation Act of 2026 (S.3741). The bill would require the Commerce Department to issue mandatory regulations on synthetic nucleic acid security — going beyond the existing voluntary framework established under H.R. 3029.

The legislation includes an explicit carveout: sequences that are “clearly non-hazardous and pose no credible threat to public health and safety” would be exempt. The design is practical: it is not a dragnet, it is a targeted checkpoint.

What the June 4 letter does is give that stalled bill a visible push. It puts the CEOs of the world’s most prominent AI companies on the record alongside the companies that would actually be regulated — and alongside scientists whose credentials are unimpeachable. Josh Wentzel, a senior fellow at the Foundation for American Innovation, told Fortune that the alignment signals something Congress should not miss: “This is bipartisan, concrete, achievable, and noncontroversial.”

That combination is rare. It should not be wasted.

What gets overlooked in the coverage

The headline writes itself — rivals agree! — but the more important story is structural.

Voluntary biosecurity screening already exists and has meaningfully reduced the likelihood of accidental or low-effort misuse. What it cannot do is hold against determined, technically sophisticated actors who understand the gaps in voluntary coverage and can route around them. Mandatory screening with recordkeeping creates a different threat environment: not impenetrable, but with full-chain accountability. Any attempt to circumvent it leaves a paper trail. “Awareness of traceability itself deters misuse,” as one biosecurity researcher put it to Science.

There is also an international dimension the letter does not fully resolve. Hayley Anne Severance of NTI Bio, whose parent organization is among the signatories, noted that for screening to be truly effective, other countries with significant synthetic DNA manufacturing capacity — including China — would need to adopt compatible rules. A U.S.-only mandate creates screening arbitrage. That is a harder diplomatic problem, and the June 4 letter is not designed to solve it. It is designed to solve the domestic chokepoint first.

The EU is moving in parallel. The European Commission’s proposed EU Biotech Act of 2025 identifies synthetic nucleic acid sequences as biotechnology products of concern and proposes a comparable EU-wide framework. The regulatory direction is converging; the timing is not yet coordinated.

Why the unified front matters beyond the specific ask

AI companies asking for regulation is still treated as a novelty. The cynical read is that incumbents want rules to raise the barrier to entry for smaller players. That read is not entirely wrong as a general phenomenon, but it misses something important here.

Bioweapons have no business case for anyone at this table. The AI companies signing this letter are not regulated entities in the synthetic biology space — they are downstream of it. They are flagging a risk to their own technology’s trajectory. An AI-enabled pandemic would be catastrophic for AI development, for public trust, and for the companies themselves. Their incentive is to prevent this, not to appear responsible.

More importantly, the specific ask is narrow enough to be credible. The letter does not call for sweeping AI regulation, model licensing, or compute thresholds. It calls for one change in one industry at one well-defined chokepoint. When a coalition this broad agrees on something this specific, the productive question is not whether they have ulterior motives. It is whether Congress will move before the capability gap makes the letter look prescient.

The window

There is always a window argument in technology policy: act now, while the solution is cheaper than the problem. Sometimes these arguments are premature; sometimes they are exactly right. The history of pandemic preparedness suggests we consistently underinvest in prevention and overpay for response.

The synthetic DNA chokepoint will not be available forever. As the technology for producing biological materials becomes more distributed and cheaper, the leverage of screening at centralized synthesis companies decreases. S.3741 is not a permanent solution to AI-enabled biosecurity risk. But it is the right tool for the moment we are actually in, backed by the people building the technology creating the risk, built into a bill with bipartisan support.

That combination of factors — the right ask, at the right time, with the right coalition — is genuinely uncommon in tech policy. What happens next is up to Congress.

If you found this useful, tap the clap button — it helps this reach readers who think seriously about AI and its second-order consequences. And if you have a take on where biosecurity policy goes from here, drop it in the comments.