AI Chiefs Urge Congress to Screen Synthetic DNA as Models Surpass PhD-Level Biology

In a show of cross-industry alignment rarely seen from competing AI companies, Anthropic CEO Dario Amodei, OpenAI CEO Sam Altman, and Microsoft AI chief Mustafa Suleyman signed a joint public letter to Congress on June 4, 2026, calling for mandatory screening of synthetic DNA and RNA orders. The letter, organized by the Foundation for American Innovation and the Institute for Progress, argues that AI systems have already crossed a threshold where they can meaningfully assist in designing dangerous pathogens, and that the commercial market for synthetic biological material offers no consistent safeguard against misuse.

The letter accompanies Senate legislation introduced by Senators Tom Cotton (R-Ark.) and Amy Klobuchar (D-Minn.), the Biosecurity Modernization and Innovation Act of 2026. The bill would require companies that sell synthetic DNA and RNA to screen both their orders and their customers against a government-maintained list of hazardous sequences and actors, with explicit exemptions for material that poses no credible public health threat. The sponsors describe it as a narrow, technically enforceable measure that fills a gap in federal biosecurity law that has existed since the commercial synthetic biology market emerged in the 2000s.

Why the Letter Carries Weight Now

The core concern the signatories raise is not theoretical. Anthropic, OpenAI, and Google DeepMind have each published internal evaluations showing that their frontier models perform at or above expert human level on a wide class of biotech questions, including questions about pathogen engineering that previously required years of specialized laboratory training to answer confidently. The letter puts the point plainly: current models "outperform PhD-level virologists" on many biotech tasks and could provide substantive assistance to someone attempting to engineer or obtain a dangerous pathogen.

What makes this moment different from earlier AI biosafety debates is the combination of capability and accessibility. The models are not behind government firewalls. They are consumer products. And the commercial market for synthetic genes and oligonucleotides, which reached roughly $3.5 billion globally in 2025, operates under a patchwork of voluntary screening standards set by individual companies, not federal law. The Biosecurity Modernization and Innovation Act would make a consistent, legally enforceable baseline universal across the industry.

"AI systems now outperform PhD-level virologists on many biotech questions and could be used to make deadly pathogens more accessible than ever before." Joint letter to Congress, signed by Amodei, Altman, Suleyman and dozens of experts, June 4, 2026

What the Proposed Law Would Do

The bill's structure is deliberately narrow. It does not attempt to regulate AI model outputs directly, a technically difficult problem with serious free-speech implications. Instead it targets the physical supply chain: the companies that synthesize and ship biological sequences on demand. Any provider selling synthetic DNA or RNA above a defined threshold of hazard classification would be required to run customer and order screening before fulfillment. Orders flagged by screening would be held pending review; customers who appear on federal exclusion lists would be blocked entirely.

Exemptions are built in for academic research that falls outside hazard categories, for agricultural and industrial applications that pose no pathogen risk, and for orders that involve only clearly non-hazardous sequences. The goal is to avoid disrupting legitimate life-science research, which depends heavily on quick and reliable access to synthetic DNA, while closing the specific gap that concerns the signatories: the ability of a bad actor to combine AI-generated design guidance with commercially available synthesis to produce something dangerous.

A Rare Show of Industry Unity

For Anthropic, the letter is consistent with a pattern of public engagement on dual-use risk. Dario Amodei has spoken at length about biological threats as one of the scenarios he considers most consequential, and Anthropic's published research on long-run AI risks lists bioweapons alongside autonomous AI as a top-tier concern. The company's Responsible Scaling Policy treats biosecurity capability as a trigger threshold, meaning models that cross a certain bar of demonstrated bioweapon-relevant capability face mandatory additional safety review before deployment.

What is less typical is the cross-company coordination. OpenAI and Anthropic are direct competitors in every material sense. So are Microsoft AI and Anthropic, given that Microsoft has invested heavily in OpenAI while Anthropic counts Google and Amazon as strategic investors. For the three firms to co-sign the same letter to Congress, with their names appearing alongside those of dozens of biosecurity researchers and national security officials, signals that biosecurity represents a category where commercial competition recedes. Whether that alignment extends to the legislative process, where tech industry lobbying positions often diverge sharply from public statements, will become clearer as the bill moves through committee.

The Biosecurity Modernization and Innovation Act has bipartisan support from its sponsors, which improves its odds in the current Senate. But the bill will face technical objections from synthetic biology industry groups, which have argued that mandatory screening could slow fulfillment times and disadvantage domestic suppliers relative to foreign competitors who face no equivalent requirements. The letter's signatories appear to anticipate this objection: the joint statement explicitly acknowledges the commercial impact and frames the screening cost as a necessary and proportionate biosecurity investment, not a compliance burden.

For the broader question of how AI companies should engage with government on the risks their own products enable, the letter sets a model. Rather than waiting for regulators to act, or lobbying defensively when legislation appears, the CEOs of the three leading AI labs chose to name a specific risk, endorse a specific legislative solution, and add their signatures to a public call for action. Whether that approach becomes a template for future AI policy engagement, or remains an isolated episode tied to the particular urgency of biosecurity, is an open question as the industry prepares for its next wave of capability releases.