Shomei is two things in one product: a fast, deterministic memory engine your agents recall from, and the governance layer that makes every one of those memories inspectable, enforceable, and provable. You don’t bolt compliance onto memory later; the same graph serves both.
The same memory layer that serves agents also answers access, restriction, hold, export, and erasure requests. Each control is scoped, receipted, and built to be checked later.
Subject access reports with source-linked memory and a fulfillment receipt.
Remove a subject from covered stores and record the boundary of what was erased.
Preserve the record while preventing ordinary AI recall until policy changes.
Freeze destruction, remove from ordinary use, and receipt every touch.
Track governed exports and record external deletion as verified or residual risk.
Detect covered memory that reappears through restore, migration, or integration drift.
Encrypted storage in your own cloud, you hold the keys, with no dependency on a particular vector backend, a small data-processing footprint and a tractable security review.
Deletion and suppression lists are ingested from the systems you already run, privacy-rights platforms, CRM, CDP, so “who must be forgotten” reconciles itself rather than becoming a manual list.
The natural first step produces the subject-access and exposure picture before any destructive action is enabled. See your exposure, then turn on governed deletion.
Deploy read-only in your own cloud and get a subject-access and exposure map in days, before you change a thing.
Governed, provable AI memory. Deletion that stays deleted, across your AI, with proof.