Detection is the wrong question.
Can students show their work?
AI detection tools produce probability scores. Probability scores produce disputes. Disputes produce lawsuits. Signet /Draft gives your students a way to prove their own authorship — before any question is raised, with evidence they own, verified by a third party.
The academic integrity landscape has split into three approaches.
How do you preserve trust in student work when AI tools are now part of the writing environment?
AI Detection.
* Probabilistic AI analyzing finished work. *
Scan a finished document. Produce a probability score. Decide.
The detection model has well-documented problems: false positive rates that disproportionately flag non-native English speakers, neurodivergent students, and formally trained writers. Lawsuits from falsely accused students. An arms race between detectors and humanizer tools that has no finish line. Major institutions — Vanderbilt, Michigan State, UT Austin — have disabled these tools entirely.
Stanford - Vanderbilt - Michigan - Humanizers
Writing Surveillance.
* Detection disguised as process recording. *
Observe students while they work. Surface the data to faculty.
A different approach, but the institution's exposure increases rather than decreases. These tools require invasive permissions to read document content, transmit it to vendor servers, and in some cases apply the same probabilistic classification they claim to replace — just earlier in the process. The student has no artifact, no agency, and may not know they're being observed. The institution inherits a new data relationship: student work flowing through a third-party vendor, stored on vendor infrastructure, governed by vendor terms.
Process Evidence.
*Student-controlled records of how the work developed.*
The student documents their own writing process and submits a verified record alongside their work. The institution receives evidence the student created — not a score a vendor produced.
Process evidence is not a keystroke replay or a behavioral surveillance feed. It's a structured record that helps a student show how their work developed, without asking software to decide what happened. No AI classification. No probability scores. No risk assessments. No new data relationships with vendors — student content never leaves the student's device.
Signet /Draft is the product built for this category. The institution becomes the trusted holder of authorship records its students generated, not a vendor of accusations.
Reduced exposure. Stronger evidence. No new risk.
No false positive liability.
Signet /Draft does not classify student writing as AI-generated or human-generated. It does not produce a probability score or an accusation. Instead, it gives students a process record they choose to submit, leaving academic judgment with the institution.
FERPA-clean by architecture.
Student work never touches Signet's infrastructure. No student papers stored. No perpetual content license. No third-party vendor holding educational records.
Narrower compliance scope.
Signet does not process student data on behalf of the institution. The student's relationship with Signet is direct. When a Data Processing Agreement applies, the scope is meaningfully reduced: student content never reaches Signet's infrastructure, so the data flows that typically drive vendor compliance obligations are absent by design.
The evidentiary position shifts.
When a student submits a Signet /Draft certificate alongside their paper, the institution is reviewing evidence the student proactively created , instead of a score a vendor produced. That is a different legal footing in any proceeding that follows.
Complements what you already have.
/Draft fills the process gap that content analysis doesn't cover. It doesn't replace your existing tools. It gives your students a way to bring their own evidence to the table — which strengthens your position, not just theirs.
Your Google Workspace.
Their .edu identity. No IT lift.
Students sign in with the .edu Google account your institution already manages. Their name is drawn from your directory. Controlled by your IT department, not by the student, not by Signet. The certificate carries the same identity trust as your LMS, your library system, and your student email.
No new credentials. No enrollment. No IT changes. Any professor verifies from any browser by scanning the QR code or dropping the file. No account required.
The standard is shifting from "detect AI" to "prove human."
Institutions are facing litigation over AI misconduct accusations — expulsions overturned, discrimination claims filed, due process violations alleged. The legal question is whether a probability score constitutes sufficient evidence. Courts are answering.
The U.S. Copyright Office concluded in January 2025 that human creative process is the standard for copyright protection — and explicitly recommended that creators keep records of their human contributions. A Signet certificate is that record. Cryptographically signed. Independently timestamped. Verifiable by anyone.
Institutions that encourage students to document their own authorship are building a compliance posture aligned with where the law is heading.
What Signet /Draft is — and is not.
Signet surfaces the data. We don't score it, flag it, or interpret it. Your institution's academic integrity standards , along with your policies, your faculty's judgment, and your established processes — All remain exactly where they belong. With you.
Bring it to your students.
We're working with writing programs, writing centers, and academic integrity offices who want to see what this looks like in practice. If you're interested in piloting Signet /Draft with your students — at no cost — we'd like to understand your context and get you set up.
Or reach us directly: hello@getsignet.app