How AI Assistants are Moving the Security Goalposts

This Krebs on Security article warns that autonomous AI assistants (notably OpenClaw) are introducing substantial new attack surfaces: exposed admin interfaces and credentials, prompt-injection and supply-chain attacks that can install rogue agents, and AI-augmented threat actors able to scale compromises (e.g., an AWS-documented campaign affecting 600+ FortiGate devices). The piece describes real incidents and demonstrations where attackers or misconfigurations enable credential theft, message injection, large-scale automation of attacks, and easier lateral movement, and it urges isolation, strict configuration, and new defensive approaches against agent fragility.