AI EncyclopediaAISafety & rules
Security vulnerability
보안 취약점
A security hole in a program that a bad actor can use to get in, steal information or damage a system.
Overview
People make programs, so mistakes creep in. A mistake that someone else can use to get in without permission or to seize control is called a vulnerability. One that the maker does not know about yet, or has not released a fixed version (a patch) for, is especially dangerous because there is no good way to block it. It is called a zero-day.
AI has become quite good at reading huge amounts of code and finding vulnerabilities. Defenders can use it to find and close holes first, but attackers can use the same tools. That makes it even more important to install patches quickly when they come out.
An everyday example
It is like a gap somewhere in your garden wall. If the owner finds and blocks it first, nothing happens. If a thief finds it first, locking the front door does no good.
Why it matters
News of AI finding many vulnerabilities at once is becoming more common. This is why you should not put off update notices on your phone and computer.
Also calledvulnerability, security hole