Privacy · 5 min read
Any AI agent that books flights, orders food, and shops for you has to see a lot. The privacy question is not just what it sees in the moment, but what it keeps, what it learns, and whether the version of you it builds can ever leave.
As of late August 2026, Meta has not published Hatch's final privacy terms. We do not know exactly how long inferred preferences will be retained, whether they will be used to train models, or how complete a deletion will be. Meta's track record with other AI products suggests user content can be used for training unless a user opts out in an eligible region.
Privacy is not only about who can read your messages. It is about who owns the model of you that gets built from them. An agent inside a platform that already monetizes attention and behavior is incentivized to make that model more useful to the platform over time. That is not a conspiracy; it is the business model.
A memory vault inverts that. You are the customer, not the product. The vault's incentive is to keep your data safe, portable, and under your control — because that is the only thing it sells.
Konshus does not train on your memory, does not sell it, and builds every feature around the idea that you can leave with your data at any time.
Meet KonshusKonshus is one way to solve this — a persistent memory vault and portable persona that follows you across ChatGPT, Claude, Gemini, and whatever ships next.
Meet Konshus