Inside each app, mostly yes. ChatGPT lets you delete individual memories or turn memory off, Claude and Gemini have their own controls, and account deletion escalates from there. The uncomfortable part is that you are always asking the company to forget, one app at a time, and you can never verify it happened.
In ChatGPT: Settings, Personalization, Memory, then Manage, where you can remove single memories or wipe all of them. In Claude and Gemini the controls live in similar settings pages. These remove what the product shows you. What you cannot see is whether derived data, caches, or training artifacts kept any shape of you. You take their word for it.
The problem compounds. Every assistant, agent, and AI-powered app you touch builds its own copy of you. Deleting yourself becomes a part-time job across a dozen settings pages, and the next app starts collecting the day you try it. You do not have a delete problem, you have an ownership problem.
The pupil model flips it. AIs do not each keep their own profile of you; they read your one file, with your consent, while you allow it. When you revoke, every AI loses access at once, and the revocation is verifiable offline, so you do not have to trust anyone’s word. Watch a live AI lose access mid conversation in the sixty-second demo.
Take the free four-minute read, get your word and your one-of-a-kind eye, and add your pupil to ChatGPT or Claude in about two minutes. Revoke it all with one tap, anytime.
Take your read, free Watch the revoke demoIt is gone from the memory store the product uses to personalize your chats. Whether any trace persists in derived systems is not something you can inspect, which is exactly the trust gap a provable revocation closes.
Any AI holding a Pupul token can check a signed status list and see the grant is dead without calling Pupul. The token still looks valid cryptographically, but access is provably revoked.