AI and Biosecurity Risks Outpace Federal Regulatory Response, CRS Report Warns
核心洞察
A Congressional Research Service report highlights that AI tools may lower barriers for developing biological threats, yet no comprehensive federal legislation governing AI development has been enacted.
Executive Order 14292 issued in May 2025 directed revision of dual-use research and nucleic acid synthesis screening policies, but as of July 2026 neither had been updated.
Industry has moved ahead with self-governance, including OpenAI (搜索)'s biosecurity framework, Anthropic (搜索)'s proposed policy, and an open letter from researchers calling for mandatory nucleic acid synthesis screening.
The convergence of artificial intelligence and biotechnology is creating new biosecurity vulnerabilities that federal regulators have been slow to address, according to a recent Congressional Research Service (CRS) report. While AI is already being used to design DNA sequences, characterize proteins, discover drugs, and power autonomous laboratories, some researchers argue these same tools could be repurposed to produce biological and chemical compounds of concern—potentially democratizing access to knowledge that historically required specialized expertise.
The report underscores a regulatory gap: no federal legislation establishing broad regulatory authorities for AI development and use has been enacted, and key policy revisions mandated by the White House remain incomplete more than a year after their deadlines.
Executive Order 14292 and Unmet Deadlines
On May 5, 2025, the Biden administration issued Executive Order 14292, "Improving the Safety and Security of Biological Research." The order directed the Office of Science and Technology Policy (搜索) (OSTP) to revise or replace the 2024 United States Government Policy for Oversight of Dual Use Research of Concern and Pathogens with Enhanced Pandemic Potential within 120 days, and to revise or replace the 2024 Framework for Nucleic Acid Synthesis Screening within 90 days.
As of July 2026, the CRS has not been able to determine whether either policy has been revised or replaced. The 2024 Framework for Nucleic Acid Synthesis Screening, which established requirements for recipients of federal funding for research related to synthetic nucleic acids, had not been updated despite the 90-day deadline.
Industry Self-Governance Moves Faster
In the absence of comprehensive federal legislation, industry stakeholders have developed their own standards and frameworks. In May 2026, OpenAI (搜索) published a governance framework covering biosecurity risks and launched competitions offering up to $50,000 for identifying jailbreaks of ChatGPT's biosafety safeguards. Anthropic (搜索) followed in June 2026 with a proposed policy framework recommending that frontier AI developers test models for catastrophic risks, publish findings, and be accountable to a government agency.
Also in June 2026, a coalition of life sciences researchers, AI companies, and biotechnology developers—including members of the International Gene Synthesis Consortium (搜索) (IGSC)—issued an open letter to Congress calling for mandatory screening of synthetic nucleic acid orders. The letter placed responsibility on DNA synthesis companies rather than on AI model developers or biological design tool providers.
The IGSC, an industry-led group of gene synthesis companies, has already developed "a common protocol to screen both the sequences of synthetic gene purchase orders and the customers who place them," implemented through industry standards and requirements for federally funded research.
Skepticism Tempers Alarm
Not all experts believe current AI capabilities warrant heightened concern. Nobel Prize-winning chemist Jennifer Doudna (搜索) stated in a June 2026 interview that "biology is complex" and "innovation is still really in the domain of human beings right now. … I'm not seeing chatbots coming up with a brand-new idea." The CRS report notes that AI-generated biological designs would still need to be physically synthesized and evaluated to determine practical utility, including assessment of predicted biosafety or biosecurity risks.
Nevertheless, the National Institute of Standards and Technology (搜索) (NIST) has warned that AI-designed novel DNA sequences of concern could be undetectable by current sequence screening tools. NIST's Biosecurity for Synthetic Nucleic Acid Sequences effort partners with stakeholders to improve DNA screening standards and evaluate mitigation opportunities for risks associated with AI biodesign tools.
Congressional Activity
Several bills introduced in the 119th Congress address aspects of the AI-biotechnology convergence. S. 4363, the Engineering Biology Readiness Act, would require the next biennial biodefense threat assessment to include analysis of engineering biology risks and recommendations to modernize biosafety, biosecurity, and biodefense authorities. S. 3952, the Future of Artificial Intelligence Innovation Act of 2026, establishes standards and evaluation tools for AI R&D and requires examination of safeguards against unintended use of AI for developing chemical, biological, radiological, nuclear, and explosive threats. H.R. 3029, the Nucleic Acid Standards for Biosecurity Act, would direct federal agencies to carry out measurement research supporting best practices for nucleic acid synthesis biosecurity.
The CRS report concludes that these bills, individually and collectively, could support the advancements of AI and biology while also providing some safeguards. Congress may continue to assess whether current federal policies and oversight are sufficient or whether new regulatory authorities are needed to address potential biosecurity concerns at the intersection of AI and the biological sciences.