### AI Lock-In Raises New Questions About Control

AI lock-in is becoming a central issue as enterprises decide which models, platforms and AI stacks they will trust. In this episode of The Open Current, Alan Shimel and Margaret Dawson examine whether AI is creating another public-cloud-style lock-in cycle. The discussion connects open source, sovereignty, data control and vendor influence into one larger enterprise technology debate.

Margaret argues that AI lock-in is hard to separate from data ownership. Many AI platforms are still black boxes. Enterprises may not fully know where their data goes, how it is used, or whether it becomes part of future model training. That uncertainty makes control and choice more important as AI moves deeper into business workflows.

### Open Source and Sovereignty Meet the AI Stack

The conversation also explores the role of open source in AI. Alan and Margaret discuss concerns around model marketplaces, open-weight models, proprietary platforms and the possibility of major vendors controlling important access points. They compare today’s AI market to earlier technology waves, including public cloud, GitHub, Red Hat and Kubernetes.

Sovereignty is another recurring theme. The episode looks at data sovereignty, AI sovereignty and the need for organizations to understand where their technology runs. For global enterprises, those issues affect compliance, risk and long-term flexibility.

### Composability Becomes the Alternative

The Open Current episode also looks ahead to KubeCon and the collision of Kubernetes, AI workloads and sovereignty. Margaret notes that AI workloads are containerized workloads. That makes Kubernetes and cloud native platforms part of the larger AI lock-in conversation.

The larger takeaway is that AI lock-in is not only about one model. It is about the full stack around the model. That includes agents, protocols, security controls, data access and operational tooling.

For technology leaders, the path forward is open composability. Enterprises need components that can be swapped, integrated and governed across vendors. Without that flexibility, AI systems could become another generation of black-box platforms that are difficult to leave once they become embedded.