October 2026 Summaries

A startup team tested an autonomous pentesting agent called Morpheus against its development environment and found that Agent0, an AI-based SRE helper with access to logs and code, could detect and investigate the activity despite not being designed as a security product. The experiment emerged informally through internal discussion and focused on using Agent0 to identify attack patterns, create monitoring rules and dashboards, send Slack alerts, and answer investigation questions through chat. Initial detection relied on simple anomaly signals such as bursts of client errors followed by a successful request, while later monitoring was extended to production and authenticated attack scenarios after checking existing controls, log schemas, and baseline activity. The account argues that increasingly capable attacker agents make continuous testing and rapid alert-to-action workflows more important than periodic penetration tests alone, while emphasizing that Agent0 is an additional security layer rather than a replacement for SOC, EDR, or other established defenses. Its broader message is that AI can lower the cost and effort of rapidly prototyping useful tools, allowing teams to turn exploratory ideas into iterative operational controls.