What happens when AI begins to design viruses?
Researchers at Stanford University and the Arc Institute used a genome-generating AI model called Evo, built on the same class of architecture as large language models but trained on genetic sequence data instead of text, to design functional bacteriophage (bacteria-infecting virus) genomes from scratch.
The team generated hundreds of thousands of candidate genome sequences computationally, selected around 300 for physical synthesis, and successfully produced 16 complete, functioning bacteriophages capable of killing drug-resistant strains of E. coli bacteria.
This is described as the first peer-reviewed demonstration of a generative AI model designing an entire functional viral genome that was then physically synthesised and shown to work, published in a leading scientific journal.
Biosecurity researchers, including specialists associated with Johns Hopkins University, have warned that existing safeguards for screening synthetic DNA orders are built to detect sequences matching known dangerous pathogens, and cannot reliably flag novel AI-generated sequences that differ substantially from anything previously catalogued.
The demonstration targeted only bacteriophages (which do not infect humans), not human pathogens; the concern raised is about the underlying capability and the governance gap, not about an immediate human health threat.
Dual-Use Research of Concern (DURC) and Biosecurity Screening
Dual-use research of concern refers to legitimate scientific research that, because of its nature, could be misapplied to pose a significant biological threat to public health, agriculture, or security. Since the mid-2000s, gene-synthesis companies in the US and Europe have voluntarily screened customer orders against databases of known hazardous sequences before fulfilling them, a practice later formalised through industry bodies such as the International Gene Synthesis Consortium (IGSC, formed around 2009), which focuses on sequence screening, customer screening, recordkeeping, and reporting suspicious orders. This screening model works by pattern-matching against a list of known dangerous sequences.
Key Details
- DURC concept underpins US and international biosafety oversight frameworks for gene synthesis and gain-of-function research.
- International Gene Synthesis Consortium (IGSC): industry self-regulation body, formed circa 2009.
- Screening approach is signature/database-matching based, not designed for novel AI-generated sequences.
The Stanford/Arc Institute demonstration exposes a structural weakness in this screening model: because AI-generated genomes can diverge significantly from any previously known sequence, they may not trigger existing pattern-matching filters, creating a governance blind spot precisely in the DURC framework meant to catch such risks.
Biological Weapons Convention (BWC) and Article IV Obligations
The Biological Weapons Convention (opened for signature in 1972, entered into force in 1975) is the first multilateral treaty banning an entire category of weapons — it prohibits the development, production, stockpiling, and acquisition of biological and toxin weapons. Article IV of the BWC obligates state parties to take necessary national measures, including legislative steps, to prohibit and prevent the development of biological weapons within their territory, which is the legal basis for national biosafety and biosecurity regulation, including oversight of DNA synthesis. India is a state party to the BWC. Critics note the treaty lacks a formal verification or inspection mechanism, unlike the Chemical Weapons Convention's OPCW.
Key Details
- BWC: opened for signature 1972, in force 1975; first treaty to ban an entire weapons category.
- Article IV: requires domestic legislative/regulatory measures against bioweapon development.
- No standing verification body (unlike OPCW for chemical weapons) — a long-flagged structural gap.
AI-generated pathogen design sits squarely within the "necessary measures" states are meant to legislate for under BWC Article IV; the emergence of AI-assisted genome design adds urgency to long-standing calls to modernise the treaty and national biosafety regulations, since the 1972-era treaty framework did not anticipate generative-AI-designed biological sequences.
India's Biosafety and Biotechnology Regulatory Framework
In India, biosafety regulation for genetically engineered organisms and associated research operates primarily through the Department of Biotechnology (DBT) guidelines and the Rules for Manufacture, Use, Import, Export and Storage of Hazardous Microorganisms/Genetically Engineered Organisms or Cells, notified under the Environment (Protection) Act, 1986. Oversight bodies include the Genetic Engineering Appraisal Committee (GEAC) under the Ministry of Environment, Forest and Climate Change, and Institutional Biosafety Committees (IBSCs) at the institution level. India is also a party to the Cartagena Protocol on Biosafety (a supplementary agreement to the Convention on Biological Diversity, in force since 2003), which governs the transboundary movement of living modified organisms.
Key Details
- Rules notified under the Environment (Protection) Act, 1986 govern hazardous microorganisms and genetically engineered organisms in India.
- GEAC (under MoEFCC): apex regulatory body for genetically engineered organisms.
- Cartagena Protocol on Biosafety: in force since 2003; India is a party.
As AI-assisted genome design capability spreads globally, India's existing GEAC/IBSC oversight structure, designed mainly for genetically modified organisms and recombinant DNA research, will face pressure to extend explicit coverage to AI-designed synthetic biology outputs, an area current rules do not squarely address.
- AI model used: Evo (genome language model family developed by Stanford/Arc Institute researchers).
- Candidate genomes generated computationally: several hundred thousand.
- Candidates advanced to physical synthesis: approximately 300.
- Functional bacteriophages produced: 16, effective against drug-resistant E. coli.
- Biological Weapons Convention: opened for signature 1972, entered into force 1975.
- India's Cartagena Protocol on Biosafety membership: in force since 2003.
- Target organisms in the study: bacteriophages (non-human-infecting viruses) — not human pathogens.