Florin Pățan

Many companies responded to this by either banning AI, buying a Data Loss Prevention (DLP) product, or signing an enterprise deal with an expensive vendor. However, bans get ignored, DLP doesn't understand prompts, and enterprise deals means you live with whatever models and rules that vendor locks you into.

At ArdanLabs we went a different direction and started building Kuzco. We believe this is a better approach to solving this and other related problems.

In this talk, I'll go over the Kuzco architecture, explain why the usual answers fall short, and what we're doing differently. Such as:

- Each user gets an isolated AI machine that only has the data and tools that user is allowed to access.
- All model and tool calls go through a single gateway, so you can control and report on every model interaction.
- Companies should have ownership of the source code and the ability to customize the platform for their data needs.