AI Policy
AI supports a controlled editorial workflow. It does not decide what is true, what is approved or what is published.
Permitted uses
Approved uses include grouping monitored headlines into story leads, suggesting categories or tags, summarising editor-selected sources, and preparing a structured first draft from a human brief.
The current drafting system uses a documented model and prompt version. Generation records retain the requesting editor, selected evidence, warnings and the resulting draft revision.
Human oversight
An administrator must verify citations, edit the copy, review uncertainty notes, record approval against a specific revision and separately publish it. Server-side controls prevent an unapproved draft from appearing in public queries.
Data boundaries
Sensitive tax identifiers, personal accounts, employee finances, personal debts, insurance details and other private material are excluded from public profiles and AI prompts. The internal AI inventory records the system purpose, model, data categories, oversight, retention and risk classification.