AI-designed proteins expose gaps in DNA biosecurity screening
Researchers led by Microsoft published a study in Science demonstrating that AI can design proteins mimicking dangerous biological threats that may evade existing biosecurity screening practices. The study involved a consortium including Twist Bioscience and industry experts, with restricted data access managed by the International Biosecurity and Biosafety Initiative for Science.
What happened
Researchers led by Microsoft published a peer-reviewed study in Science titled 'Strengthening nucleic acid biosecurity screening against generative protein design tools,' demonstrating that AI can design proteins that mimic the activity of known biological threats.
The study was conducted as a red-teaming exercise by a consortium of scientific and industry experts including Twist Bioscience and the International Gene Synthesis Consortium. Current biosecurity and screening practices work well for known proteins and sequences, but the research highlights vulnerabilities as AI capabilities evolve.
Researchers and the journal restricted access to study data and software, partnering with the International Biosecurity and Biosafety Initiative for Science to vet access based on legitimate need.
Reported impact
AI's ability to design hazardous biological threats lowers the barrier to creating possibly harmful agents. The research identifies unforeseen vulnerabilities in biosecurity screening that will require future corrections.
Experts emphasize the need for further lab work to verify if AI-designed proteins actually function as predicted in practice, citing international treaties prohibiting biological weapons advancement.
Index analysis
The study demonstrates that as AI capabilities evolve, biosecurity screening practices must evolve just as quickly to prevent potential misuse. The work is described as proactive risk assessment to get ahead of rapid development in the field.
Current industry best practices for biosecurity screening are robust and highly effective for known proteins and sequences, but emerging AI-driven protein design tools represent a new frontier requiring novel screening approaches.
Governance controls
Twist Bioscience has developed and advanced stringent biosecurity screening methods and standards in parallel with commercial offerings. The company demonstrated proactive leadership by participating in this consortium initiative to identify industry next steps for responsible innovation.
The researchers implemented restricted data access managed by the International Biosecurity and Biosafety Initiative for Science, which vets access based on legitimate need. This novel approach aims to manage risks associated with sharing potentially dangerous information.
Classification
Sources and evidence
Sources not yet published
This record was researched and written by the Index from original reporting. Its sources are still being verified by an editor, and they are listed here once that check is complete. Until then the record carries no reviewer.