What makes a personal AI agent private?

A personal AI agent sees the most sensitive parts of life: money, plans, contacts, routines, and private notes. Privacy cannot be a paragraph in a policy; it has to shape the product architecture.

Useful protections include workspace isolation, encrypted sensitive fields, minimal retention of raw uploads, export controls, and deletion paths that include records, derived summaries, and long-term memory.

The goal is simple: the assistant should be helpful because it remembers what matters, while still giving you clear ownership and control over the data behind that memory.