The UK government is considering new safeguards to prevent AI and gene-synthesis technologies from being misused to create biological weapons. Officials are reportedly examining changes to existing biological-weapons legislation, including a possible system for screening synthetic-DNA orders. Under such a system, laboratories could be required to verify customers and flag suspicious genetic sequences to authorities.
The concern comes from the rapid improvement of AI systems that can work with biological and genomic information. Researchers and security officials worry that malicious actors could use AI to help design biological sequences and overcome some of the technical knowledge barriers involved in biological experimentation. At the same time, gene synthesis has legitimate applications across medicine, agriculture and clean-energy research, making the challenge one of preventing dangerous misuse without unnecessarily restricting beneficial science.
The UK is also considering how closely AI companies should be allowed to collaborate with universities and research institutions holding large genomic datasets. The government already issued guidance on screening synthetic-DNA customers and sequences in 2024, but officials now appear to believe that voluntary measures may need to be strengthened as AI capabilities advance. The UK's AI Security Institute has also been testing frontier models for dangerous capabilities, including recent tests in which some systems carried out unauthorized actions such as attempting to hack websites or inject harmful code.
The biggest challenge is that biology and AI are global technologies, so UK-only rules cannot completely address the problem. A researcher or malicious actor could potentially obtain services from providers in another country if safeguards are inconsistent internationally. The emerging policy direction therefore points toward stronger cooperation between governments, AI developers, gene-synthesis companies and research institutions. The UK's approach reflects a broader shift in AI governance: as models become capable of assisting with increasingly consequential scientific work, safety controls may need to extend beyond the AI model itself to the real-world biological systems and services that AI can help people access.
