“

Recent developments involving frontier AI models have further demonstrated the ability of advanced systems to circumvent safeguards, interact with external systems, and substantially enhance offensive cyber capabilities. … These incidents underscore the need for the U.S. government to better understand and address emerging threats before they result in more serious consequences. … We must also continue to improve our understanding of these increasingly complex systems, whose internal processes can be difficult even for experts to fully interpret, predict, and control. [Young, in a letter asking National Security Adviser Marco Rubio to have the National Security Council convene recurring talks with AI developers on AI security threats.]