Three hours. One stolen password. An entire cloud environment compromised.

Anthropic's recent threat intelligence report highlights a cyberattack where AI agents allegedly automated significant portions of the compromise. This included finding exposed credentials, accessing company systems, and extracting data. In one instance, over 2,100 login tokens were reportedly collected from more than 40 company accounts within approximately 34 hours. This raises concerns for security professionals regarding the tracking of active API/AI keys across various projects, applications, repositories, and third-party systems.