Key facts
- AI models can design novel viruses, potentially as effectively as nature.
- The viruses created in a recent experiment infect bacteria, not humans.
- Advancements in AI capabilities could make viral bioweapons easier to design and produce.
- Biological attacks using novel viruses could be transmissible and spark pandemics.
- Mandatory gene synthesis screening for DNA/RNA sales and equipment is gaining traction.
- AI companies' CEOs endorsed gene synthesis screening, and federal legislation is proposed.
Recent experiments have shown that artificial intelligence models are capable of designing novel viruses, a development that has prompted concern among biosecurity experts about the potential for creating biological weapons. While the viruses created in one such experiment infect bacteria rather than humans, the advancement highlights the growing risk that AI could be misused.
Experts note that while current AI chatbots may not be sophisticated enough to assist lone actors in creating deadly pathogens, newer models trained on extensive biological data could pose a greater threat. The ability to design transmissible viruses could lead to biological attacks that spread rapidly, potentially sparking pandemics, unlike past attacks involving substances like anthrax or ricin which were not human-to-human transmissible.
To mitigate these risks, a strategy gaining momentum involves mandatory gene synthesis screening. This would require companies selling synthesized DNA or RNA, as well as gene synthesis equipment, to verify customer legitimacy, record all orders for traceability, and screen against known dangerous genetic sequences. The goal is to eventually detect novel biological threats not yet on any list. CEOs of leading AI companies have endorsed this approach, and federal legislation is being considered to make it harder to obtain the necessary materials for biological weapons.