Who owns an agent’s memory?

Persistent assistants turn memory into an institutional question. A remembered preference can be useful; a remembered confidential conversation can be a liability. The longer an agent works across tasks, the less adequate it is to treat memory as a private convenience.

Users need to know what is retained, who can inspect it, how long it survives and how it can be corrected or deleted. Organizations need boundaries between teams and purposes. Memory is not neutral context. It is an archive with permissions.