Abnormal AI Deploys Amazon Bedrock AgentCore for Scalable Email Security

Abnormal AI, a behavioral security service protecting more than 25 percent of the Fortune 500, has put Amazon Bedrock AgentCore Code Interpreter into production for its real‑time inline email threat detection. The agents process billions of messages daily, executing code in secure, serverless MicroVM sandboxes to block threats before they reach inboxes.

The Code Interpreter service offers configurable session lifetimes, isolated host‑level sandboxes, flexible networking, and pre‑loaded Python and Node.js runtimes. Abnormal AI integrates it as the compute scratch pad in the third tier of a three‑tier detection pipeline—handling tens of thousands of the hardest cases after lightweight heuristics and medium‑size ML models filter the bulk traffic.

The deployment illustrates how agents benefit from a dedicated compute environment for tasks such as counting, data visualization, and code verification, and provides practical guidance on sandbox security, programmatic verification, and checkpointing for long‑running jobs.