Frontier AI labs are stepping up biological risk testing, which is harder than cybersecurity testing, where capabilities can be tested in digital environments (Financial Times)

Frontier AI labs are increasing their focus on biological risk testing, a process deemed more challenging than cybersecurity testing. This difficulty arises because cybersecurity capabilities can be evaluated in digital environments, whereas biological risks involve more complex, real-world implications. Executives and biosecurity experts are concerned that advanced AI models could potentially be used to facilitate the creation of novel viruses or bioweapons, highlighting the critical need for robust biological risk assessments.