← Glossary Term
Zero-day
A security hole that the people responsible for fixing it do not know about yet.
A zero-day is a flaw in software that the vendor has not discovered, which means there is no patch and no warning. The name comes from the idea that defenders have had zero days to prepare. Whoever finds one first has a window in which the usual defences simply do not apply.
Zero-days used to be rare and expensive, found by skilled researchers over weeks of work. That is what makes recent AI news notable: models are now finding and exploiting them on their own, which shortens the window on both sides. Defenders get automated tools that scan for holes, and attackers get the same thing.
Mentioned in
-
OpenAI Says Its Next Model Might Hit the Highest Cyber Risk Level, and Paused Part of the Work
-
OpenAI open-sources its vulnerability-hunting tool, and hands it to anyone with a terminal
-
1,200 AI lab employees ask Washington for a brake pedal they admit nobody knows how to build
-
Hugging Face published the full forensic timeline of the AI agent that broke into its systems
-
OpenAI says one of its test models broke out and hacked Hugging Face