Consent is the only moat left in data

Every few weeks another model company is sued over where its training data came from. The arguments differ in detail and converge on one question: was this yours to take?

That question has an uncomfortable property. It cannot be answered retroactively. A dataset assembled by crawling has no record of consent because no consent was sought, and no amount of subsequent documentation can manufacture one. The provenance either exists from the moment of collection or it does not exist at all.

What this means commercially

For anyone buying research data, the risk has moved. It used to sit with the quality of the data. It now sits with its origin. A dataset that cannot demonstrate lawful collection is a liability with a spreadsheet attached, however clean the rows look.

This is why we built Surdatics the way we did. Not because consent is virtuous — though it is — but because it is the part of the supply chain that cannot be retrofitted, which makes it the part worth owning.

What "we asked" has to mean

Saying data was collected with consent is easy. Three things make it checkable:

The person knew what they were agreeing to. The purpose is shown before the first question, not buried in a document nobody opens.

The person was paid. Compensation is funded into escrow before a survey goes live, so payment does not depend on the researcher still being around when the work is done.

The record cannot be revised. Every response carries an account of what was asked, when it was answered, and what was paid. A digest of that record is anchored so it cannot be quietly edited later.

That last point is the one that does the work. A claim you can change after the fact is a marketing line. A claim you cannot change is evidence.

The limit, stated plainly

None of this makes a dataset risk-free. Anonymisation has limits: a small dataset with rare demographic combinations can narrow to an individual however the columns are labelled. We flag those at listing time rather than implying a guarantee nobody in this industry can actually make.

What it does give you is the ability to answer the question that is increasingly being asked in court — where did this come from — with something more than an assurance.