Asuu

How do you delete from a record that others have relied on?

The right to leave and a trustworthy audit trail pull in opposite directions. We do not yet know where the line belongs.

Open question. Not yet tested.

On this page

Why it is hard.

Two things we have promised pull against each other. The charter says a person can leave and ask for their record to be deleted. The Network exists so that nobody can quietly rewrite what a model did and who approved it. A record that can be erased is not a trustworthy trail. A trail that can never be erased is not yours.

The conflict is practical. A clinician made a decision using information that the patient later asks to remove. A year on, the decision is questioned. What should the clinician be able to show?

What has been tried.

The whitepaper gives one answer. It keeps two layers apart. The clinical record stays inside the institution and can be corrected or erased under the law of that place. The shared ledger holds only fingerprints of what existed at each moment, with a marker added when something is corrected or removed. Industry has a related trick, in which data is encrypted and the key is destroyed.

A direction we would test.

We would add two things. The first is a visible scar. Where something was removed, the record says so: an entry existed here, and it was deleted at the person's request on this date. It does not say what the entry was. Anyone who relied on it is told that it has gone.

The second is a choice between erasing and sealing. A sealed entry is locked under a key held by an independent advocate and can be opened only by a court. The person chooses, having been told plainly what each option means for them and for the people who treated them.

How we would know we were wrong.

A fingerprint of a short record can sometimes be reversed by guessing the contents and comparing. If we cannot make that impractical, the scar becomes a leak, and what looked like deletion was nothing of the kind. We would ask outside cryptographers to try before any patient is asked to trust it.

Who could solve it.

A cryptographer willing to attack the idea of a scar before we rely on it. A data protection lawyer who can say what must be kept and what may go, country by country. A clinician who has been asked to defend a decision years later, and knows what they needed to have in their hands.

Where this connects.

Solve one with us.

Tell us which problem you work on, and what we have got wrong about it.