How do you detect coordinated activity across multiple AI agents?
Correlate identity, activity, resources, timing, destinations, credentials and behavior across agents into a single incident graph. Several agents that each stay under individual thresholds can show a combined pattern, such as five agents sharing one credential and one destination, that becomes one incident for investigation. Correlation is evidence, not proof of intent.
- Seven correlation dimensions.
- 5 agents → 1 correlated incident in the demo.
- Human review before irreversible action.
Frequently asked questions
Does correlation prove the agents are malicious?
No. It links evidence so analysts can investigate. Many correlated patterns have benign explanations.