At 50 to 1,000 employees, your company ships AI features weekly, indexes internal documents and customer records into RAG pipelines, and holds its most valuable IP in fine-tuning datasets. Enterprise customer security reviews, SOC 2 audits, and investor diligence now ask how you govern all of it, usually before you have a dedicated AI governance team.

This dinner brings together CTOs, VPs of engineering, and security leaders from venture-funded Bay Area technology companies to share what they put in place first. Guests discuss how to find unsanctioned AI tools across distributed teams, keep vector databases and training data from leaking, and inspect prompts and outputs in real time without slowing developers down.

The discussion centers on AWS network controls that give you enforcement where policy documents fall short, including AWS Network Firewall for outbound traffic to external model APIs, AWS PrivateLink for private access to Amazon Bedrock, and Amazon Bedrock Guardrails for sensitive information filtering.