The story’s score and claims

Comments on No. 86 of 123 · developers.googleblog.com · edition of Saturday 14 November 2026

Agent Anomaly Detection scores on benchmarks we didn't write

Summary

Google tested its Agent Anomaly Detection feature on five public, non-Google agent-security benchmarks covering prompt injection and direct harmful requests. The feature flagged 87% of harmful instructions that agents actually carried out, with 2% to 16% false alarms on benign sessions.

Comments

Ranked by a public rubric, not by votes: what a comment adds counts, not who wrote it. How comments are ranked · Comment rules