Understand how MOSS detects when your AI agents change behavior over time.
MOSS learns your agents normal behavior and alerts when it changes.
AI agents can drift: a customer support bot starts accessing financial data, a summarization agent begins sending data externally, or a trading agent increases its transaction frequency. MOSS builds a behavioral fingerprint for each agent and monitors for drift from the baseline.
Drift Detection
Gradual behavior change over time. Agent fingerprint shifts from baseline. Alerts when stability score drops below threshold.
Anomaly Detection
Sudden behavior change. Agent does something completely unexpected. Immediate alert, no gradual buildup.