Proof of personhood
for education
for education
OriginStory for education
Technology developed at Arizona State University that establishes a live human, the student they claim to be, was present and responsible for the work at the moment it was made.
Start here
Choose an author on the right. Watch what happens to the work on the left.
The Colorado River Compact allocated more water than the river reliably carries, a shortfall masked for decades by unusually wet years and by states declining to draw their full entitlement. Arizona's junior priority under the 1968 authorization means it absorbs shortages first, which is why the state's groundwater has quietly become a buffer rather than a reserve. Any durable settlement has to price that buffer, not simply reallocate the surface flow.
Three different students. One artifact. The grade book records the same A for all three, and nothing in the submission tells the university which one it just credentialed.
For a century, captured work was evidence that something happened. That assumption is gone, and it went quietly, inside every course at once.
What is actually at stake
They educate students, and they vouch for what those students know and can do.
Grades, degrees and certificates are claims about one person's capability, made by the institution, to everyone who will ever rely on them. For decades the second job rested on the first. Essays, problem sets, code, projects and exams stood in as evidence of learning, and the evidence was good enough because producing it required the learning.
Generative AI severs that link. It does not make students dishonest, and most are not. It makes the artifact silent about its own origin.
Employers have already started reporting the gap between the degree a recent graduate holds and how that graduate performs in the role. The claim is being tested in public, by the people who relied on it.
The two answers on the table
Run the submission through a detector.
Detectors learn the statistical fingerprints of the models that exist today. The next model does not have those fingerprints, and neither does a student who lightly rewrites. OriginStory's own researchers published the argument for why this class of approach does not generalize.
The deeper problem is that even a perfect detector answers the wrong question. It tells you a machine was involved. It never tells you what the student knows.
Chalak, Lenz, Bliss, Liss and Berisha, "Why Speech Deepfake Detectors Won't Generalize", arXiv preprint, September 2025.
Remove AI from coursework.
Students will use these tools for the rest of their working lives, in every field the university prepares them for. An institution that trains them without the tools is training them for a job that no longer exists.
It also does not work. Enforcement moves the work off campus and out of view, and the university ends up policing the artifact instead of teaching the person.
This is an assessment problem, not a detection problem.
The term
The live human who is the student they claim to be, physically present and responsible for the work.
Not a login. Not a face scan at the start of a session. A property of the moment the work is made, that travels with the work afterward.
How presence becomes proof
A microphone and a second signal from the device the student is already using, read together at the moment of speaking.
The signals are matched locally. The underlying biometric data does not need to travel to a central database.
What leaves is not a body. It is a cryptographically verifiable record that this person was present and responsible.
The proof survives the medium. It stays attached through a recording, a transcript, a submitted file or a live session.
Inside a course
Where it applies
An essay, a problem set, a block of code, a recorded submission. The proof attaches at the moment of creation and travels with the file wherever it goes afterward.
A session, an oral check, a defence of an argument. The proof holds continuously while the student is speaking, rather than being taken once at the door and assumed for the rest.
Command, the third, governs devices and instructions. It sits outside the classroom, which is why it is not on this page.
Verification is not a gate the student passes once at the door. It is a property of the moment, and it stays visible where the work actually happens.
What it does not do
OriginStory does not judge whether a student has mastered a concept, and it does not try to detect whether AI was used. Faculty continue to define mastery and design the assessment. This is a personhood layer underneath the assessment, not a grading system on top of it.
From assessment to credentials
Select a link to see what it carries, and what fails without it.
A verifiable record that a specific live person was present and responsible, established at the moment of the work rather than asserted afterward.
Everything downstream. The demonstration and the credential are both statements about a person the institution cannot identify.
Held as a chain rather than a line on a transcript, the record travels. A graduate keeps verifiable evidence of what they actually demonstrated, and the institution that awarded the credential keeps its meaning intact.
The opportunity
Establishing identity and assessing an individual has traditionally meant a controlled testing room or paid remote proctoring.
Both are expensive. Both are hard to scale. Both are poorly matched to students who learn continuously, with AI, wherever they happen to be, which is the population that has grown fastest.
A trusted personhood layer inverts the model. Students use AI throughout the learning process, and the university keeps secure moments to verify individual understanding along the way. The check moves inside the learning instead of sitting outside it in a locked room.
For an institution operating at ASU's scale, online learners included, that makes it possible to widen access to programs while demonstrating exactly what the credential still represents. The assessment problem grows more acute as education becomes more digital, more personalized and more widely available, which is the direction ASU chose a long time ago.
Beyond the degree
The same student becomes a job candidate.
Résumés, portfolios, coding exercises, written responses and interviews can all now be generated or mediated by AI. Employers are facing their own version of the assessment problem, and they are facing it about your graduates.
What they want to know is whether the person on the other side of the conversation is a Trusted Human, and whether that person holds the capabilities their record represents. A connection established during education carries into hiring. That is worth something to the graduate, and it is worth more to the university whose name is on the degree.
Where this comes from
That is the whole argument. Everything above is how it is done.