Anomaly Detection: definition and example
Anomaly detection flags activity that departs from an expected pattern without matching a known bad signature. In AI runtime security it applies to sequences of prompts and tool calls as well as single events. It finds new attack forms but produces more uncertain alerts than rule matching.
Facts read on .
Example
Straiker says Defend AI traces every user, model, tool-chain and agent-to-agent interaction to support root-cause analysis, anomaly detection and audit readiness.
Sources
- Straiker page (straiker.ai), read