The compliance story
AI is already on campus. Virgil makes it visible, governable, and defensible.
Students are using AI either way. The question for a campus is whether that use is hidden in a browser tab or happening in the open, under the course's own rules, with a record everyone can stand behind. Virgil is a desktop app that does real academic work in real files, and it was built so that the professor's policy, not the student's memory, is what governs each session.
Three people have to say yes. Each gets a straight answer.
For faculty
Your policy, enforced
Write the AI policy once. Virgil holds every student session to it and keeps the record.
For IT and review
Little to review
A desktop app that does not hold student data. The vendor review is short on purpose.
For the institution
A defensible position
AI use in the open, under course rules, producing graduates who can direct AI professionally.
For faculty · Academic integrity
Most AI tools ask students to remember the rules. Virgil enforces them.
The course AI policy loads into each class project, straight from the syllabus. Different courses, different rules; each project knows its own. PSYCH 240 allows AI research help, ENGLISH 225 doesn't.
Compliance barrier
Require explicit approval before each session starts.
Policies · ENGL225_syllabus_AI_policy.pdf
· Never write final prose. Outlines and feedback only.
· Every source gets a real citation I can verify.
· Flag anything that needs my own analysis.
Virgil refuses clear violations and flags work that gets close to a line. Every approval is a receipt, logged to audit.jsonl inside the project.
For faculty · The record
Professors get receipts, not guesswork.
A read-only link shows the live project: drafts, sources, the board. Process is visible, not just the final PDF. The session log separates the student's work from the AI's, so integrity conversations start from a record instead of an accusation.
Students leave your course knowing how to direct AI professionally, not just how to prompt around the rules.
For IT and the review board · Data custody
Student work never moves into our cloud.
Virgil is a desktop app. The vendor review is short because there is very little to review: we do not hold the data.
| Student work, drafts, sources | Plain files on the student's own machine, never a vendor database |
| Model keys | Controlled by the student or the institution, scoped per project |
| AI reasoning and chat | The student's or school's own model account, under API terms: not used to train anyone's model |
| Optional Engine credits | Routed through zero-data-retention endpoints |
| Session and audit log | audit.jsonl, inside the project folder |
| Virgil's servers | Only what a team explicitly chooses to sync for group work |
For the institution
A position you can defend to anyone.
Virgil does not hold student records or coursework by default, which keeps data-protection exposure where a university wants it: minimal. There is no scraped training data, no shadow profile of the student, and no lock-in; work is ordinary files that stay with the student after graduation.
Model-agnostic by design, Virgil runs on whichever provider your campus already has terms with, so existing agreements with Anthropic, OpenAI, Google, or Microsoft keep doing their job.
And students graduate having practiced directing AI inside rules, the working norm every employer now expects.
How a pilot works
A pilot fits inside one course. No procurement mountain.
Week zero
Pick one course. The professor writes the AI policy into the syllabus; we help draft it if there isn't one yet.
Day one
Students install Virgil free and drop the syllabus in. The policy becomes each project's rulebook automatically.
All semester
Virgil enforces the policy session by session. Faculty get read-only project links and session logs whenever they want them.
Semester's end
The review board gets what it never had before: a record of how AI was actually used, course by course, receipt by receipt.
The packet
Two PDFs for the people who ask good questions.
One page · PDF
The compliance one-pager
Integrity enforcement, data custody, and why the vendor review is short. Built for IT and review boards.
Download →
16 slides · PDF
Virgil for Academia
The full walkthrough: a semester in Virgil, from syllabus drop to defensible final paper. Built for faculty.
Download →
Bring AI on campus into the open.
We work directly with faculty and review boards on pilots, course policies, and vendor questions. One course is enough to start.