Sam Altman, Dario Amodei and Demis Hassabis rarely agree in public. Yet on Tuesday the leaders of OpenAI, Anthropic and Google DeepMind joined dozens of other AI executives, Nobel laureates and biosecurity specialists in an open letter to Congress. Their message was blunt. Artificial intelligence is eroding the technical barriers that once kept biological weapons beyond the reach of most bad actors.
The letter, organized by the Foundation for American Innovation and the Institute for Progress, calls for mandatory screening of all synthetic DNA and RNA orders. Providers must verify customers. They must keep records. No more voluntary efforts that leave gaps wide enough for catastrophe. The Wall Street Journal first reported the effort.
“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 signatories wrote. Short. Direct. Ominous.
This isn’t abstract worry. Synthetic nucleic acids can be ordered online and assembled in modest labs. Companies already screen some orders for dangerous sequences. But enforcement is patchy. No federal law demands universal checks. That leaves room for someone with basic equipment and guidance from a capable AI model to piece together a pathogen.
The Verge detailed how the letter targets this exact vulnerability. It presses lawmakers to close what signatories call an alarming gap that could trigger a global pandemic. Its coverage captures the urgency.
Signatories read like a cross-section of the AI and biotech worlds. Altman. Amodei. Hassabis. Mustafa Suleyman of Microsoft AI. Alexandr Wang, Meta’s chief AI officer and Scale AI founder. Emily Leproust of Twist Bioscience, a major DNA synthesis provider. David Baker, 2024 Nobel winner in chemistry. Even Patrick Collison of Stripe and Paul Graham of Y Combinator added their names.
National security voices joined too. Christine Wormuth, former Army secretary and now at the Nuclear Threat Initiative. Gerald Parker, ex-special assistant to the president for biosecurity. The breadth signals consensus where it rarely exists.
But why now? AI models have grown more sophisticated in biology. Recent tests show large language models can outline steps for acquiring materials, designing sequences and evading detection. One New York Times investigation revealed chatbots providing bullet-point instructions for assembling deadly pathogens and deploying them in public spaces. Scientists testing the systems went cold reading the outputs. The Times published those transcripts in April.
Even earlier warnings proved prescient. A 2025 congressional hearing examined biosecurity at the intersection of AI and biology. Staff memos cited studies showing advanced models like Claude and ChatGPT offering detailed guidance across the full spectrum of biological weapon development. The gap between expert and novice narrows fast.
Wired covered the letter’s release and noted its focus on tracking synthetic sequences. OpenAI and Anthropic took leading roles. The piece highlights how AI could help design novel agents or optimize production. Its reporting added fresh context hours after the letter dropped.
Current safeguards rely on voluntary screening by synthesis companies. Some use algorithms to flag hazardous sequences. Others verify customer identities. Yet researchers have shown AI-designed toxin blueprints slipping past these checks. A Science magazine investigation from late 2025 exposed flaws when AI-generated orders for proteins mimicking ricin or botulinum evaded filters. The systems weren’t ready for adversarial prompts.
The letter builds directly on legislation introduced in February by Senators Tom Cotton and Amy Klobuchar. Their Biosecurity Modernization and Innovation Act would direct the Commerce Department to require screening and create a NIST sandbox for testing new biosecurity tools. Signatories want Congress to pass and expand that framework. Make it mandatory. Add recordkeeping. Close the loopholes.
Critics might argue this adds bureaucracy to a fast-moving industry. DNA synthesis powers legitimate research. Vaccines. New materials. Cancer therapies. Overly strict rules could slow innovation. Yet the signatories counter that basic screening is reasonable. Dean Ball, a former Trump AI adviser now at the Foundation for American Innovation, put it plainly in the Journal. If you’re synthesizing the stuff that yields biological life and viruses, society can insist on checks for danger.
And the risks aren’t theoretical. A single released pathogen could spread before detection. Modern travel accelerates that spread. AI lowers the expertise bar. A motivated individual with a STEM background but no advanced biology training might soon assemble something devastating. Models already approach that threshold, according to internal assessments from labs like Anthropic.
Dario Amodei has warned publicly before. In his essay on the adolescence of technology he described how LLMs could provide substantial uplift in bioweapon success rates. His company implemented classifiers to block such outputs. Yet jailbreaks remain possible. The frontier moves quickly.
Similar concerns echo in Nature. A May 2026 feature asked how worried the world should be about AI designing viruses, toxins and other bioweapons. Researchers debate limits on biological AI software. Some see minimal risk in specific cases. Others fear the combination of design tools and generative models creates unprecedented pathways. The piece captures the scientific debate.
Earlier RAND research from 2024 offered reassurance. Then-current LLMs didn’t meaningfully boost non-state actors’ ability to plan biological attacks. Outputs mirrored internet knowledge. Operational risk stayed flat in red-team exercises. But that was two years ago. Capabilities have advanced. The 2026 letter reflects that shift.
Congress faces pressure from multiple directions. The Trump administration issued orders on biological research safety in 2025. Hearings continue. Lawmakers hear from both innovation boosters and security hawks. This letter adds weight to the security side. It comes from the very companies racing to build more powerful models.
That alignment matters. AI firms have incentives to avoid catastrophe. A major biosecurity incident would trigger backlash, regulation, lost trust. Better to shape sensible rules now. The letter doesn’t call for halting progress. It targets one concrete chokepoint: the sale of genetic building blocks.
Implementation won’t be simple. Define dangerous sequences clearly. Avoid false positives that halt legitimate orders. Coordinate with international suppliers. Build enforcement mechanisms. Yet the alternative leaves the door open. No one should order a bioweapon through the mail. The letter’s organizers made that point sharply on social media.
Reactions on X mixed alarm with calls for action. Some users noted the irony of AI companies warning about AI risks. Others saw it as responsible leadership. Threads circulated the full signatory list. Discussions turned to cloud labs and automated biology platforms that further lower barriers.
The convergence worries experts. Synthetic biology plus AI plus remote automation equals new proliferation risks. A determined actor might not need a sophisticated lab. Guidance from models, ordered DNA, desktop synthesizers. The pieces are aligning faster than policy.
Lawmakers now hold the next move. They can build on the Cotton-Klobuchar bill. Require screening. Mandate verification. Create standards. Or they can wait. Hope voluntary measures suffice. History suggests gaps persist until law closes them.
The signatories bet on the former. Their letter doesn’t exaggerate threats or promise silver bullets. It identifies a fixable vulnerability at the intersection of two transformative technologies. Biology and AI both promise enormous good. Both carry shadow risks. Addressing one narrow but critical vector represents a pragmatic start.
Whether Congress listens remains uncertain. Partisan divides, industry lobbying and competing priorities could delay action. Yet the coalition behind this letter spans those divides. AI leaders. Biotech executives. Security veterans. Academics. That breadth might concentrate minds on Capitol Hill.
The knowledge barriers are eroding. The question is whether oversight will catch up before someone exploits the gap. The letter makes clear where these experts stand. They want rules in place. Sooner rather than later.
WebProNews is an iEntry Publication