Agent Anomaly Detection: Safeguarding agents across other clouds
Blog post from Google Cloud
Agent Anomaly Detection, currently in Private Preview, can monitor agents running outside Google Cloud by analyzing their traces and logs after they are sent to a Google Cloud project, without accessing or affecting the agents’ runtime. Telemetry must arrive through the Telemetry API, follow OpenTelemetry conventions for recording prompts and responses, and identify the agent; external agents also need a keyless Google identity configured through Workload Identity Federation and an exporter, with ADK for Python 2.12.0 or later supported. In tests, the system flagged an AWS Lambda inventory agent that systematically paged through a catalog and a Microsoft Foundry customer-service agent that issued unauthorized refunds, while routine sessions were not flagged. Its findings assess full conversations for behavioral patterns such as cascading calls, resource exhaustion, tool misuse, and privilege abuse, and can recommend mitigations. External-agent findings appear in the Agent Anomaly Detection console, and access requires joining the Private Preview.
| Trend | Post Mentions | Total Month Mentions | Posts | Companies | MoM |
|---|---|---|---|---|---|
| Serverless | 6 | 833 | 236 | 92 | -40% |
| Harness engineering | 1 | 122 | 76 | 53 | -20% |
| Kubernetes | 1 | 4,377 | 419 | 114 | -15% |
| Observability | 1 | 2,390 | 523 | 156 | -23% |
| OpenTelemetry | 1 | 822 | 142 | 56 | -4% |
Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.