Abnormal AI says it runs agent code for inline email threat detection

Abnormal AI says it has deployed Amazon Bedrock AgentCore Code Interpreter in production for real-time inline email threat detection. The company says its systems process billions of email messages daily, while inline agents handle tens of thousands of the hardest cases by dynamically writing analysis scripts in isolated sandboxes. Abnormal says it uses a no-egress sandbox configuration to limit threat-intelligence data exfiltration. AWS identifies Abnormal as a customer and says the views are the customer's.