Researchers from Johns Hopkins University, the University of Oxford, Stanford University, Columbia University and New York University are calling for safeguards against certain infectious disease data sets that could allow artificial intelligence to design deadly viruses. The warning highlights a seriously overlooked security flaw in the development of artificial intelligence.
The researchers point out that once high-risk biological data spreads on the open web, it can no longer be taken back. This means that even if relevant regulatory regulations are introduced in the future, if the dangerous knowledge itself has been widely disseminated, any regulatory measures will be ineffective. This irreversibility makes the management and control of biosafety data particularly urgent and complex.
Currently, with the rapid development of artificial intelligence technology, especially its increasingly in-depth application in the biomedical field, the ability of AI systems to acquire and process high-risk biological data is also constantly increasing. Researchers worry that if appropriate access restrictions and security measures are not taken to these sensitive data sets, malicious actors may use AI technology to engineer or modify pathogens to create highly contagious or lethal biological threats.
The call reflects deep academic concerns about the risks of dual-use AI. While open sharing of scientific data is critical to advancing medical research and public health, certain potentially dangerous biological information can have catastrophic consequences if it falls into the wrong hands. Therefore, how to find a balance between promoting scientific progress and ensuring biological safety has become a key issue that needs to be solved in the field of AI governance.