← AI security in plain English

AML.T0012

MITRE ATLAS / Valid Accounts

Adversaries obtain and abuse the credentials of existing accounts to gain initial access. Legitimate credentials blend into normal activity and often bypass controls aimed at detecting intrusion.

Think of it likeWalking in the front door with a key that works, so nobody thinks to question you.

In plain English

The quietest way into an AI platform is a real login. Stolen or leaked credentials produce activity that looks authorized, because as far as the system is concerned it is.