A zero-day is a software flaw that attackers exploit before the vendor has released a fix. The name reflects that defenders have had zero days to prepare a patch. The term is often applied both to the vulnerability and to an attack that uses it.
Zero-days are hard to defend against because no official fix exists when exploitation begins. Defenders rely on other measures in the meantime, such as vendor-provided workarounds, restricting access to the affected service, extra monitoring and virtual patching through other controls. Once the vendor releases a patch, the flaw is generally no longer a zero-day, although systems that have not applied the fix remain just as exposed.
The term is sometimes used loosely. A vulnerability that is newly published is not necessarily a zero-day, and one that is described as a zero-day is not necessarily being used widely. Exploitation evidence, and who reported it, matters more than the label.
When a zero-day affecting your software is announced, read the vendor's advisory for mitigations, apply them, check whether your exposure is real by confirming versions and configuration, and watch for the patch. Review logs for activity matching the published indicators, and apply the fix as soon as it is available.