Zero-click refers to an attack technique that can compromise a device or system without any action by the user, such as clicking a link, opening a file, or installing software. It exploits vulnerabilities in processes that software performs automatically, such as receiving a message or rendering a preview, so unlike phishing, which requires the victim to be tricked into doing something, even careful users have little defense.
Typical examples include spyware infections that exploit flaws in smartphone messaging apps or image-processing routines. In the AI field, the term is applied to prompt injection attacks in which an AI assistant or agent automatically ingests an email, document, or web page containing hidden instructions and acts on them. Because the agent processes the content on its own, data can be leaked without the user doing anything.