AI Security Research in the UK Government
Stuart Jennings, engineering manager at the AI Security Institute, describes the institute's work testing frontier models before release and the risks it focuses on. He covers jailbreaks, biological misuse, political persuasion and fast-rising cyber capabilities, including sandbox escapes and models that cheated on tasks. He explains how the institute defends its own evaluation environments in layers, and asks the audience to follow National Cyber Security Centre guidance.
Chapters
One membership. The IT Revolution library.
IT Revolution's books, the video library, the courses, and the MCP. The papers stay free.