Microsoft Researchers Uncover AI-Generated "Zero Day" Threats in Biosecurity Systems
A team of researchers at Microsoft has successfully used artificial intelligence to create "zero day" threats in biosecurity systems designed to prevent the misuse of DNA, highlighting the dual-use potential of AI algorithms. According to a study published today in the journal Science, the team's findings demonstrate how generative AI can be exploited to bypass existing protections and potentially manufacture deadly toxins or pathogens.
The researchers, led by Microsoft's chief scientist Eric Horvitz, used generative AI algorithms that propose new protein shapes to discover vulnerabilities in biosecurity systems. These types of programs are already being used in the hunt for new drugs at well-funded startups like Generate Biomedicines and Isomorphic Labs, a spinout of Google. However, as Microsoft's researchers have shown, these same algorithms can be used to create "zero day" threats, which are previously unknown vulnerabilities that can be exploited by attackers.
The breakthrough raises concerns about the misuse of AI in bioterrorism and underscores the need for more robust safeguards against adversarial AI attacks. According to Horvitz, "our research highlights the potential risks associated with the use of generative AI algorithms in biosecurity systems." He emphasized that the goal of the study was not to create a threat but rather to identify vulnerabilities that could be exploited by malicious actors.
The researchers' findings have significant implications for the development and deployment of AI-powered biosecurity systems. As the study's lead author, Horvitz noted, "we need to develop more robust safeguards against adversarial AI attacks, including the use of techniques such as differential privacy and secure multi-party computation." These measures can help prevent the misuse of AI in bioterrorism and ensure that these powerful technologies are used for beneficial purposes.
The study's publication comes at a time when concerns about the misuse of AI in bioterrorism are growing. In recent years, there have been several high-profile incidents involving the use of AI-generated biological agents, including a 2020 incident in which a group of researchers used AI to design a novel pathogen that could evade existing defenses.
The Microsoft team's research highlights the need for greater collaboration and coordination between governments, industry leaders, and academic institutions to address the risks associated with AI-powered bioterrorism. As Horvitz noted, "we must work together to develop more robust safeguards against adversarial AI attacks and ensure that these powerful technologies are used for beneficial purposes."
In response to the study's findings, a spokesperson for the US Department of Homeland Security stated that the agency is "aware of the potential risks associated with the use of AI in bioterrorism" and is working to develop more robust safeguards against adversarial AI attacks. The spokesperson emphasized that the agency is committed to protecting public health and safety from emerging threats.
The Microsoft team's research has significant implications for the development and deployment of AI-powered biosecurity systems, and highlights the need for greater collaboration and coordination between governments, industry leaders, and academic institutions to address the risks associated with AI-powered bioterrorism. As Horvitz noted, "we must work together to develop more robust safeguards against adversarial AI attacks and ensure that these powerful technologies are used for beneficial purposes."
This story was compiled from reports by MIT Technology Review and MIT Technology Review.
