Treat versioning as an assignment problem
AI functions, learning activities and teaching decisions evolve. Without a preserved version, staff and students may be unable to establish which permission, restriction or documentation expectation applied to a particular piece of work. A “last updated” date is insufficient when it replaces the earlier text. The useful question is precise: which published rule was intended and accessible for which cohort, assignment and working period?
Start by separating layers of authority. University-wide policy, examination regulations, programme material, a course statement and an assignment-specific note may have different owners. A course convenor cannot quietly amend a higher-level requirement merely by issuing a fresh version number. The matrix therefore links to authoritative documents, records a possible conflict and names the role able to resolve it.
Distinguish editorial fixes from substantive changes
A spelling correction which leaves meaning intact is different from a new prohibition, a broader permission or an additional evidence requirement. Classify the revision and display the affected old and new wording. This makes it possible to detect when something described as a clarification actually changes what students are expected to do. Sequential numbers are convenient, but clarity and consistent use matter more than a particular numbering convention.
A substantive change should have a recorded educational or operational reason. A revised learning outcome, an ambiguous example, an unsuitable data route or a redesigned assessment may justify reconsideration. “A new AI product exists” is not, by itself, a complete rationale. The entry should explain which activity is affected and why the previous wording no longer describes the intended learning or assessment conditions.
A durable policy generally describes activities rather than maintaining an exhaustive list of brands. Brainstorming, literature discovery, translation, text generation, coding support and language correction remain understandable when product names change. Where a particular platform must be identified for privacy, licence or access reasons, preserve enough information about its version, configuration and approved route to make the decision intelligible.
Plan the effective point and communication together
A publication timestamp does not demonstrate that the relevant students received the change. Place the version and a concise explanation in the institution’s established course channel. The assignment brief, marking rubric and policy should cross-reference the same identifier. A seminar announcement can help explain the revision, but it should not be the only durable record of a rule affecting assessed work.
Changes during an active working period require particular care. The authorised institutional role must determine whether the earlier version continues, an option is offered or a new requirement can properly take effect. The template does not assume the answer. Its purpose is to make existing cases and the selected transition visible instead of treating retrospective expectations as self-evident.
Test comprehensibility with realistic questions. May a tool suggest search terms? May it improve the language of a student-authored paragraph? What disclosure is expected? Ask people who did not write the policy to classify these scenarios from the text alone. Different reasonable answers indicate a drafting gap. Record and correct that gap rather than blaming readers or claiming an unsupported comprehension rate.
Keep earlier versions readable and unchanged
The archive stores every published version with its identifier, scope, publication date and effective point. The current policy provides a visible route to earlier versions. A cryptic filename or a history visible only to administrators may not allow an affected person to reconstruct the applicable rule. Stable links and accessible formats are part of the publication process.
Do not silently rewrite an archived policy. If a broken link needs repair, log the technical correction without altering the historical rule. Personal questions or individual cases do not belong in a public changelog. Policy history and case records have different purposes and should follow their respective privacy and retention controls.
The Higher education practice hub connects other implementation topics. The guide to building a course AI policy supports the content of each version. Exactly one foundation destination is included: academic integrity foundations.
Run a consistency check before release
- The version, status and responsible role are unambiguous.
- The changelog shows old and new wording with a specific rationale.
- Every current assignment maps to an identified version.
- The competent role has resolved transitional treatment.
- Brief, rubric and policy carry the same identifier.
- Earlier versions remain readable and discoverable.
- The publication channel and question route are named.
- No unapproved institutional rule is presented as fact.
After publication, check that links, downloads and version labels work. Repeated questions can inform the next draft, but an answer to one student does not automatically alter the published policy. A documented decision, new version and established communication step are what turn interpretation into a traceable change.
Review should also cover dependencies. If the disclosure form names fields removed by the new policy, or the rubric still rewards a superseded practice, the policy number alone will not prevent confusion. Record each linked artefact and its owner in the release check. Unresolved dependencies should be stated openly until the authorised correction is available.