Feeling overwhelmed by AI? Learn the Centaur Mindset. Read More →
A terminal from the lab, condensed. With ADK_MODEL=gemini-2.5-flash, five alerts including an SSH brute force come back FP 0.3 with 'Failed to parse LLM response: True positives require exactly three remediation steps', and an alert with an injected SSH banner comes back FP 0.97 citing the embedded triage note. Below, the raw model replies for the same brute-force alert read TP 0.95 with 5 steps and TP 0.98 with 4 steps.

It Said Threat, It Stored False Alarm: Stress-Testing an AI Alert Triage Copilot

A security operations center (SOC, the team that watches an organization’s security alerts and responds to them) drowns in alerts. Every intrusion detection rule, endpoint agent, and cloud monitor writes them, and most turn out to be harmless: the vulnerability scanner doing its weekly sweep, Windows Defender fetching updates, an admin running a script that happens to look odd. Somebody still has to look at each one and decide. That decision is called triage, and the two answers are a true positive (a real threat that needs action) or a false positive (a false alarm that can be closed). ...

November 20, 2026 · 35 min · Scott Algatt