Abnormal AI Uses Amazon Bedrock to Secure Billions of Emails

Abnormal AI, protecting over 25% of Fortune 500 companies, has deployed Amazon Bedrock AgentCore Code Interpreter to power real-time inline email threat detection at billion-message scale. The system uses a three-tiered architecture, with AI agents handling the most complex cases that would otherwise require human analysts. Amazon Bedrock AgentCore Code Interpreter provides ephemeral, serverless MicroVM sandboxes that give agents a compute environment for data analysis, code execution, and verification beyond pure language reasoning. Abnormal AI also reports that 80% of its own code changes now involve AI agents, with 40% built entirely end-to-end by background agents.