Remember when your Trusted Compute Base was deterministic? When policies were static and clearly specified? When "allow" meant "allow"? When your compute topology made sense and fit neatly on a whiteboard? When "policy as code" meant having one policy engine instead of dozens? Traditional policy as code approaches often rely on assumptions that are increasingly outdated, and as these assumptions get broken, tooling and processes get left behind. The gaps between having an implemented policy and automating a controlled process are large and closing them feels nearly impossible when the requirements tighten faster than the improvements can be rolled out. And then there's the "Guardrails for AI" question: lots of answers, but no leading solutions. We will discuss what it takes to implement and scale Guardrails, why most initiatives fall short, and why the policy engine is only one of the many necessary capabilities. In this session, Hazel will draw upon her experience of addressing these issues and will provide insights for leaders on ways to prepare their organisations for the coming challenges ahead.
Hazel spends her days building out teams of humans as well as the systems and tooling to make life better for others. She currently serves as a Fellow of the Nivenly Foundation, and is fondly known as the Infrastructure Witch of Hachyderm (a popular Mastodon instance). One of her favourite joys is the light on someone's face when they understand something for the first time. She also loves shooting pool, swing dancing, and boardgames.
