AI agents could partner to identify and exploit vulnerabilities, elevate privileges, and covertly pilfer data from enterprise systems through aggressive prompting that emphasized the urgency of task completion, The Register reports.
The vulnerabilities, including three RCE flaws (CVE-2026-21666, CVE-2026-21667, and CVE-2026-21669) and one allowing execution as the postgres user (CVE-2026-21708), enable low-privileged users to execute remote code on vulnerable servers with low complexity.