Human authorship needs a definition before it needs a seal
Many labels, no definition
Machines create now. That is not in dispute, and it is not going to reverse.
What is in dispute is what human authorship is still worth, and, more urgently, what it even means. Multiple organisations are already developing competing definitions and labels for human-created work. Almost none of them agree on the criteria. Some require an audit; others let you download the badge for free. The result is a signal that means several different things at once, which is another way of saying it means nothing.
The problem underneath is that almost all serious creative work already uses machines somewhere. The writer runs a spellchecker. The musician quantises. The illustrator fills a background. A standard that draws a binary line excludes most working creators and certifies a purity that barely exists.
Provenance, not inference
Detection alone is not a sufficient foundation for establishing human authorship. Detectors can perform well under some conditions, but their results depend heavily on the model, the language, the length of the text and the domain, and paraphrasing can substantially reduce their effectiveness. Documented false positives fall hardest on non-native English writers, flagged as machines for writing carefully.
The deeper problem is not accuracy. A probabilistic inference about a finished work is not the same thing as evidence of how it was made. So we do not detect. We work on provenance: the documented history of how a work came to exist, and whether the decisions that mattered were made by a person.
That formulation admits degrees, and it should. The question is never whether a machine touched the work. The question is who decided what the work would be.
Defining that threshold, precisely enough to be verified and generously enough to include real creators, is the work. It is not a detail to be settled once a seal exists. It is the standard.
Three things this project is not
Not a token. OutAI has no cryptocurrency, no blockchain, and no relationship with any project that has used this name. Provenance here rests on governance and human audit, not on a ledger.
Not a detector. We do not analyse finished work to guess its origin. We do not sell that service and we do not believe it can be done reliably.
Not a certifier. We are not issuing seals and we are not building one. We read the criteria that others publish, we compare what they promise against what they verify, and we write about what we find. This is criticism, not competition.
What this project asks
Where does the line fall between assistance and substitution? Which creative decisions must be human for a work to be someone’s own? The standards now in use each answer differently, and none of them were built on evidence about how creative work is actually made.
This project publishes what it finds. It does not certify.