A source record
for students’ AI research.

A proposed pilot for one university course. Common Measure limits students’ AI reads to approved sources, applies licence terms and gives each student a record of what their AI read.

Approved sources.

Choose the sources the AI may read. Reads from anywhere else are refused.

Licence terms.

Publishers’ licence and AI-use terms are checked before content reaches the AI.

A record of each source.

Every source read through Common Measure, with its terms and any cost.

What faculty would see.

A record the student chooses to attach to a submission. This one is from a recorded run: four reads, two refused, each with the reason.

Scope common-measure/demo

4 mediated fetches: 2 admitted, 2 refused.

Crossing refused

Access rule: * refuse

Host outside the operator's allowed sources. No request sent.

URL
https://www.economist.com/
Reason
access rule 32 (*) refuses host www.economist.com.
Fetched
Nothing. No status, no bytes, no hash.
Read the full guide

In the record

Each source the AI read, its terms, and the AI sessions and models used.

Not in the record

The student’s typing, editing and ideas. AI tools used outside Common Measure.

No score

The record lists what happened. It gives no rating.

Who sees what in the pilot.

Security and privacy
The student
Holds the record and chooses what to share.
The university
Sees how much of its licensed content was used, by title. It does not see who read what.
Publishers
Get the usage reports their licence requires, under an identifier only they see.

What exists today and what the pilot adds

Today

  • Approved-source rules, with other reads refused
  • A record of each source, its terms and cost
  • Licence and AI-use terms checked on each read
  • Usage reports to content owners

The pilot adds

  • Records the student signs for submission
  • Library-licensed content through students’ AI tools, where the licence allows

Planned

  • Flags for retracted papers
  • Checks for references that do not exist

The pilot.

Proposed for one course or a group of volunteers.

Talk to us
Students
20 to 30. Voluntary, with no effect on grades.
Assignments
Two, with different AI-use expectations set on the AI Assessment Scale.
Review
Faculty mark the same submissions without the record, then with it.
Where it runs
On the student’s own computer, with Claude Code or Codex. A university’s own AI platform is untested.
Before any student data
Ethics review and a data agreement with the university.