AI-Assisted Research Produces Zoom Exploit in One Day with Under 20 Prompts

Security firm A Security found and exploited a Zoom screen-sharing annotation flaw (ZSB-26015) in a single day using fewer than 20 prompts directed at publicly available AI models. Researcher Idan Levcovich said producing a working exploit for this vulnerability class had previously been nation-state-level work. Zoom has patched the flaw across Windows, macOS, Linux, Android, and iOS.