Enterprise AI Reference Architecture | AI Security L69

An attacker runs code in your AI orchestrator. What can they reach? With a shared role, readable logs and open egress, almost everything. Learn security zones, identities, keys, private endpoints, egress and central policy. Includes a blast-radius lab (an illustrative model). Chapters 00:00 One compromised service 00:37 Security zones 01:58 Identity 02:52 Secrets and keys 03:19 Network and egress 03:55 Central policy 04:30 Check yourself 05:06 Summary AI Security, Responsible AI & Guardrails, lesson 69 of 83. References AWS Security Reference Architecture (AWS SRA): https://docs.aws.amazon.com/prescript... Amazon Bedrock: interface VPC endpoints (AWS PrivateLink): https://docs.aws.amazon.com/bedrock/l... NIST SP 800-207: Zero Trust Architecture: https://csrc.nist.gov/pubs/sp/800/207... Narration: synthetic voice (Kokoro TTS). Animations: Manim.