The heads of OpenAI, Anthropic and Google DeepMind are asking Congress to move biosecurity from voluntary practice to legal requirement. That matters because AI is making technical biological knowledge easier to reach, while synthetic DNA can still be ordered through unevenly screened channels.
Sam Altman, Dario Amodei and Demis Hassabis do not agree on everything, and that is precisely why this letter is worth paying attention to. The leaders of OpenAI, Anthropic and Google DeepMind have joined a wider group of AI, biotech and science figures calling on Congress to require screening and recordkeeping for synthetic nucleic acid orders, including DNA, RNA and the equipment used to make them.
This is not a broad appeal for another AI talking shop. It is a specific request aimed at a practical weakness in the life sciences supply chain. If a dangerous sequence can be designed, ordered and assembled without consistent checks, the problem is no longer just what a chatbot will say. It becomes a question of whether the physical materials needed for misuse can be stopped before they ship.
As WIRED reported, the signatories include Google DeepMind's Demis Hassabis, OpenAI's Sam Altman, Anthropic's Dario Amodei and Microsoft AI's Mustafa Suleyman, along with scientists and biosecurity specialists. Other reports also list Meta's Alexandr Wang among the supporters. The letter is titled In Support of Mandatory Nucleic Acid Synthesis Screening and Recordkeeping, and its central point is direct: voluntary safeguards are no longer enough.
The concern is not that AI has already made biological weapons easy to build. The evidence on present day risk is still mixed, and serious researchers are careful about overstating it. The concern is trend. Models are getting better at answering technical laboratory questions, helping users plan procedures and pointing them toward services that can provide biological materials.
That changes the policy problem. For years, many AI safety debates have focused on model behavior, refusals and evaluations. Those still matter, but they are only one layer. The letter pushes Congress toward a more grounded intervention: make providers check sequence orders for dangerous content, verify that customers are legitimate and keep records that can support a biosecurity investigation later.
There is a reason this proposal has attracted such a wide group. Synthetic DNA has helped accelerate vaccine research, diagnostics and basic science. Small teams can now access capabilities that once belonged only to major institutions. That is a real gain. But the same online ordering infrastructure creates a choke point where misuse can be detected, delayed or traced without shutting down legitimate science.
The industry has known this for years. The International Gene Synthesis Consortium was formed in 2009 to create voluntary safeguards for gene synthesis companies. The US government has also issued guidance through agencies including HHS and the White House Office of Science and Technology Policy. The gap is that guidance and voluntary screening do not create a uniform floor across the market.
Why Congress is the target
The letter asks Congress to act this session because the signatories want one national standard rather than a patchwork of state rules or uneven company policies. That is an important business point. If responsible providers screen orders and less careful competitors do not, the companies taking biosecurity seriously carry extra cost while the weakest links remain open.
For founders and operators in biotech, this is where the story becomes more than AI governance. Mandatory screening would affect procurement, customer verification, compliance systems and record retention. It would also shape how investors evaluate risk in companies building around synthetic biology, lab automation and benchtop synthesis tools.
For AI companies, the move is also strategic. Altman, Amodei and Hassabis are effectively saying that biosecurity should not be solved only by restricting frontier models. By pointing to nucleic acid screening, they are steering lawmakers toward a control point outside the model layer, one that applies to the materials and machines needed to turn biological instructions into reality.
That does not remove the responsibility from AI labs. If anything, it makes the responsibility more concrete. The major labs are acknowledging that their systems could lower knowledge barriers over time. They are also arguing that a useful response should involve the life sciences supply chain, not just AI usage policies written inside private companies.
There is a geopolitical angle as well. Biology does not respect borders, and neither does access to AI. A US rule could push domestic providers toward stronger practices, but the long term effect will depend on whether international bodies and foreign governments align around similar screening expectations. Otherwise, bad actors can shop around for weaker jurisdictions.
The practical test will be whether lawmakers can write rules that catch dangerous orders without smothering ordinary research. Screening databases need to be current. Customer checks need to be serious. Recordkeeping needs to support legitimate investigations without turning every lab purchase into a bureaucratic ordeal.
Still, this is one of the cleaner AI policy asks now on the table. It is narrow, understandable and tied to a real supply chain. If Congress moves, the result could become an early model for governing AI enabled risks through the systems around AI, not only through the models themselves. If it does not, the largest labs have now put their warning in public, and the next serious biosecurity debate will start from that record.
Also read: Mastercard is bringing stablecoins deeper into payment settlement • Anthropic turns its safety reputation into enterprise leverage • Reve 2.0 shows image generation is still open for startups
